Methods for predicting technical application properties of polymers

Abstract

The present invention refers to an apparatus (110) for predicting technical application properties for a polymer based on a digital representation of the polymer. A digital representation providing unit (111) provides a digital representation of the polymer representing polymer descriptors. The polymer descriptors represent parameters quantifying physicochemical characteristics of a subgroup of the polymer. A predictive model providing unit (112) provides a predictive model adapted to predict technical application properties of the polymer based on the digital representation, the predictive model being a data-driven model parameterized to predict technical application properties associated with the polymer based on the polymer descriptors represented by the digital representation. A characterization unit (113) determines the technical application properties based on the provided digital representation of the polymer and the predictive model. An output unit (114) provides the technical application properties.

Description

[Technical field]

[0001] FIELD OF THEINVENTION The present invention relates to a method, an apparatus and a computer program product for predicting technical application properties of polymers. Furthermore, the present invention refers to a training method, a training apparatus and a computer program for training a data-driven predictive model usable by the method, the apparatus and the computer program product for predicting technical application properties of polymers. [Background technology]

[0002] background In general, prediction of technical application properties of materials, especially polymeric materials, such as thermal insulation coefficient, hardness, or reflectance, is a challenging task with high industrial relevance. Existing models that perform such predictions are often very material-specific and / or computationally very expensive. Therefore, existing models cannot be easily applied, for example, to screen a large number of expected materials for a particular technical application property. Therefore, it is advantageous to provide predictability of technical application properties of polymers that allows robust application to new polymers and has lower computational costs. Summary of the Invention [Problem to be solved by the invention]

[0003] Summary of the Invention It is an object of the present invention to provide a method, an apparatus and a computer program product that allows accurate prediction of the technical application properties of polymers, has a lower computational cost and is robustly applicable to new polymers. In addition, it is a further object of the present invention to provide a training method, a training apparatus and a computer program product that allows to provide a predictive model that can be trained to provide good prediction accuracy by utilizing less computational resources, which can be used in the method, the apparatus and the computer program. [Means for solving the problem]

[0004] In a first aspect of the present invention, a computer-implemented method for predicting technical application properties for a polymer based on a digital representation of the polymer is provided, the method comprising: a) providing a digital representation of the polymer representing polymer descriptors, the polymer descriptors representing parameters quantifying physicochemical characteristics of a subgroup of the polymer; b) providing a predictive model adapted to predict the technical application properties of the polymer based on the digital representation, the predictive model being a data-driven model parameterized to be adapted to predict the technical application properties associated with the polymer based on the polymer descriptors represented by the digital representation; c) determining the technical application properties based on the provided digital representation of the polymer and the predictive model; and d) providing the technical application properties.

[0005] Since the predictive model is parameterized based on polymer descriptors representing parameters quantifying the physicochemical characteristics of the polymer subgroups, i.e. based on polymer descriptors derived from parameters quantifying the physicochemical characteristics of the polymer subgroups, the model can learn, for example based on training data, how the subgroups, e.g. monomers, generally behave after polymerization. Thus, the predictive model becomes more robust in predicting the technical application properties of polymers with new and differently configured subgroups, e.g. monomers, that are not included in the training data. In addition, deriving the descriptors for the complete polymer directly from the recipe or from the respective calculations, e.g. ensemble calculations, is generally computationally expensive, and the measurement of such characteristics is time-intensive and often expensive. In contrast, calculations for subgroups with only a few atoms and bonds are computationally less expensive, and furthermore, experimental measurements for the subgroups can be performed much more easily, or existing data sets can be used for small compounds that may form the subgroups. In addition, using the subgroups as a basis even makes it possible to determine the descriptors for the different subgroups in advance and to store the respective descriptors in a database. For the processing of a polymer, i.e. for the processing of a digital representation of a polymer, the respective descriptors of the subgroups of the polymer can then be easily and quickly received from the database. Since the number of technically relevant subgroups, e.g. monomers, is limited, it becomes possible to obtain the properties of the polymer in a less computationally intensive form and to save computational resources compared to models that are not based on subgroups. Furthermore, the use of subgroups as a basis makes it possible to address homopolymers, copolymers as well as polymer blends with the predictive model.

[0006] In addition, the training of the respective biodegradation model can be improved since the polymer descriptors of the subgroups of polymers, which include the physicochemical information of the polymer, for example the quantum chemical information of the polymer, such as the solubility in water and octanol or the molar mass of the polymer, are utilized. In particular, the utilization of the polymer descriptors of the subgroups allows such models to be trained with less training data, since the use of the polymer descriptors already presents the model with some of the correlation information that needs to be learned. This further allows the saving of the tests and experiments required to provide the training data.

[0007] The development of new chemical products tailored to application requirements is a major challenge in the modern chemical industry. In recent years, further requirements are often raised, for example regarding the environmental burden or safety of chemical products along their life cycle. Companies developing new polymers therefore need to invest very large resources not only in testing potential polymers for their intended application properties, but also in self-assessment and certification of the sustainability or safety of the products. The evaluation of such conditions of potential new polymers and testing, including laboratory space and equipment, is costly and time-consuming. Therefore, the technical properties of new materials need to be identified early in the development process. The proposed method of determining technical application properties disclosed herein allows for a faster and more efficient way of developing new materials. In an early phase, the respective technical application properties can be determined even before the synthesis of the polymer. This makes it possible to determine whether the polymer is suitable for market launch. This allows for a faster time to market. It also makes it possible to reduce waste production, since it is not necessary to synthesize the polymer to determine the application properties. The proposed method provides a digital twin for measuring the application properties of the polymer.

[0008] Furthermore, standard measurements and tests for the respective technical application properties are time-consuming and may involve, for example, waiting times of up to several months or even years. In particular, when developing new polymers for the respective applications, these time-consuming tests may strongly limit the development process. In this context, the present invention allows to provide results on new polymers immediately, significantly reducing the time until the results are available.

[0009] In addition, due to the large number of expected, but often insufficiently investigated, polymers potentially suitable for a particular application, today, technical product engineers given the technical task of finding a polymer that is not only suitable for a particular application but also meets the respective target properties must synthesize and test a huge amount of expected polymers, or must search through huge data sets and libraries in which potential polymers are stored, to find each polymer that can match the application. Even when using sophisticated design of experimental methods, it is still necessary to synthesize and experimentally test a very large number of expected polymers. In this regard, the above-mentioned method allows users, for example technical product engineers, to assist in automatically finding potentially suitable polymers much faster. Specifically, by using the above-mentioned method, the user only needs to synthesize and test potentially suitable polymers that are determined to be highly likely to meet the respective target properties. Thus, unnecessary synthesis and testing of polymers can be avoided. Thus, the method allows users to perform the technical task of finding a suitable polymer for a technical application quickly and more efficiently.

[0010] The method relates to a computer-implemented method and can therefore be implemented by a general-purpose or dedicated computer adapted to implement the method, for example by executing the respective computer program. The method is adapted to predict the technical application properties of the polymer based on the digital representation of the polymer, for example as the value of the application property. In particular, the technical application property can refer to any property of the polymer and / or of the substance at least partially composed of the polymer, for example the blend or mixture containing the polymer, that allows the evaluation of the technical application potential of the respective polymer provided after synthesis. Preferably, the technical application property comprises at least one of mechanical properties, optical properties, physicochemical properties, chemical properties, and biological properties. In general, the mechanical properties can refer to any of adhesion, tensile strength, stiffness, hardness, shrinkage, elongation, tear, tear strength, elastic rebound, compressibility, wear, flow, morphology, tactile properties, stress at break, elongation at break, particle size distribution, and degree of packing. Optical properties may generally include any of the following: color tint, turbidity, opacity, clarity, reflectance, appearance, absorption, scattering, color strength, cloud point, matteness, optical density, spectrum, and refractive index. Additionally, physicochemical properties may refer to any of the following: density, viscosity, K value, molar weight, dispersity, molar mass distribution, particle size distribution, solubility, partition coefficient, interfacial properties, surface tension, dispersibility, storage stability, odor, separation, solidification, electrical conductivity, electrical capacity, surface area, flow time, vapor pressure, VOC, solids content, hygroscopicity, magnetic, miscibility, thixotropy, phase transition properties, glass transition temperature, corrosion inhibition, solvent separation, aggregation, self-heating, impact sensitivity, loss on drying, reaction angle, electrostatic charge, minimum film formation temperature, and charge density. Chemical properties may refer to chemical resistance, reaction timing, demolding time, growth, hard / soft segment content, crystallinity, reaction temperature, reaction pressure, decomposition, thermal decomposition, photolysis, acidity, pK a, pH, moisture / water content, flammability, burning rate, autoignition, flash point, flammable gas production, reaction to fire, deflagration rate, residual monomer count, by-product production, degree of polymerization, salt content, temperature resistance, oxidation properties, reduction properties, reactivity, ash content, non-volatile matter content, stability, chelating capacity, calorific value, saponification value. Additionally, the biological properties may include any of biodegradability, biological resistance, toxicity, biotransformation, ecotoxicity, sensitization, bacterial count, enzyme activity, environmental distribution, bioaccumulation, and biological exposure.

[0011] In general, the technical application properties predicted by the method are based on the specific problem that has to be solved using the method and therefore also on the respective specific training with specific training data of the predictive model used. Thus, different predictive models can be provided for different technical application properties, or a predictive model can be trained to predict several technical application properties. However, the predictions in all cases follow the principles of the method as described above.

[0012] In a first step, the method includes providing a digital representation of the polymer, which represents the polymer descriptor. In particular, providing may refer to receiving the digital representation from a user's input, for example using a respective input unit. In addition, providing may refer to accessing a storage unit in which the digital representation is already stored. Furthermore, providing may also include receiving the polymer descriptor from another source, for example via a network connection, and providing the received polymer descriptor as a digital representation. In addition, representing or being associated with a polymer descriptor of a polymer is defined as being able to access information of the polymer descriptor. For example, the digital representation may directly include the polymer descriptor, for example in the form of values ​​for the respective quantities. However, the digital representation may also be a link to the respective polymer descriptor, via which the polymer descriptor can be accessed, or the digital representation may point to an identifier associated with the polymer descriptor that allows utilizing a respective look-up storage device to access the polymer descriptor. In addition, the digital representation may also point to information that allows deriving the polymer descriptor using one or more known relationships. For example, the synthetic specifications or structural formulas of the polymers may be utilized as digital representations to allow derivation of respective polymer descriptors using known chemical and physical laws and relationships.

[0013] In general, throughout the following description, reference to a parameter or a descriptor includes reference to both the respective quantity and the specific value of that quantity, unless explicitly stated otherwise. For example, a parameter being a temperature always refers to the quantity being a temperature, and also to the specific value of temperature that is set for that quantity. In most cases, the explicit value of a parameter may be different for different embodiments and applications, so the value is usually not mentioned. However, providing a parameter or a descriptor generally means providing information about a quantity, e.g., that the value is a temperature, and also the value or descriptor of that quantity itself.

[0014] Generally, it is preferred that the digital representation comprises a polymer descriptor, which represents a parameter that quantifies a physicochemical characteristic of a subgroup of the polymer. In particular, the polymer descriptor may refer to a physicochemical parameter of the polymer. In general, the physicochemical characteristic may refer to a physical and / or chemical characteristic, in particular a parameter, of the polymer. However, the digital representation may also be provided to allow the derivation of the polymer descriptor, for example by providing a representation of the polymer, from which the subgroups may be determined, and a polymer descriptor determined based on the characteristics of the determined subgroup. A subgroup refers to a part of a polymer, all subgroups of the polymer together forming a polymer. Generally, a subgroup may refer to a part of a polymer, the subgroups being linked together in a chain or network in a continuous manner to form a polymer. Preferably, a subgroup of a polymer refers to a repeating unit that describes a part of the polymer that when repeated produces a complete polymer chain. However, in some cases, a subgroup may also refer to a single part of the polymer that is not repeated, for example an end group of the polymer. In addition, it is preferred that the subgroup comprises a repeating part. For example, a subgroup of a polymer may comprise a repeating core that is also present in other subgroups, and further additional parts that are not present in other subgroups. Preferably, the subgroup refers to at least one of the polymerized monomers or oligomeric fragments. Preferably, the subgroup refers to the polymerized monomers. In this context, polymerized monomer refers to the monomer after it has been polymerized, also called "mer unit" or "mer". In particular, polymerized monomer does not refer to the monomer as present in the reaction mixture before polymerization, in particular the raw material, but to the repeating unit derived from the monomer that has been changed during or after polymerization. Thus, the subgroup descriptors determined for the polymerized monomers are different from the subgroup descriptors determined for the unreacted monomers before polymerization. In particular, the inventors have found that polymerized monomers make it possible to determine the polymer descriptors from the subgroup descriptors of the polymerized monomers, which allows for an accurate determination of the technical application properties.In addition, it has been found that determining the polymerized monomers for the digital representation of the polymer is particularly computationally inexpensive and allows the application of valid rules. In a further preferred embodiment, the subgroups, for example referring to the polymerized monomers, are provided as molecular models representing the chemical structure of the subgroups after their polymerization. Even more preferably, the molecular models of the subgroups are selected in a manner suitable for quantum chemical calculations, with regard to the number of atoms and their connectivity, which represents the properties of the subgroups in the polymer. Moreover, in addition to or instead of the molecular models of the subgroups treating the subgroups as monomeric structures, molecular models can also be used that refer to oligomer models that take into account the adjacent molecular structures of the subgroups in the polymer.

[0015] In general, if the digital representation of a polymer does not directly include a polymer descriptor, it is preferred to determine the polymer descriptor by determining a subgroup of the polymer. For example, the respective subgroups of the polymer may be determined using known methods. However, it is preferred that the determination of the subgroups of the polymer is performed according to the embodiment of the invention described below. In particular, it is preferred that the subgroups are determined such that the polarization in the bonds between the atoms of the different subgroups in the polymer is as small as possible, and preferably the bond order is as small as possible (e.g., single C-C bonds). In addition, it is preferred that the subgroup representing the polymer contains the same number of active non-hydrogen atoms as the polymer. Besides the active atoms, the subgroup may also contain further atoms, which may be ignored during the calculation of the descriptor of the subgroup. Furthermore, it is preferred that the subgroups are determined such that polymers containing moieties built by different polymerization techniques are sufficiently covered and satisfy the aforementioned conditions. An example is polyethers used as components for polyurethanes. In general, a database or archive can be generated with multiple reactions between polymer moieties, and the subgroups can be derived from the respective structures of the reactions. For example, specific chemical languages ​​such as SMILES and SMARTS can be utilized to easily derive the subgroups of the polymer. For example, a database of reaction SMARTS can be generated and then the corresponding reaction SMARTS can be selected based on the polymerization of each polymer. From the selected reaction SMARTS, the SMILES of the monomers of the polymer can then be directly derived and, for example, RDkit can be used to determine the SMILES of the subgroups, i.e., the number and connectivity of atoms, from the SMILES of the monomers.

[0016] The determined subgroups of polymers are associated with subgroup descriptors representing parameters quantifying the physicochemical characteristics, i.e. physical and / or chemical characteristics, of the subgroups of polymers. In particular, if the polymer descriptors are not directly provided by the digital representation, it is preferred to determine the polymer descriptors by determining a respective subgroup descriptor for each subgroup and then determining the polymer descriptor based on the subgroup descriptors of the subgroups, for example by averaging. Thus, the method preferably comprises, for a polymer, first providing or determining a subgroup from the digital representation of the polymer, then determining or providing subgroup descriptors, i.e. values ​​of parameters quantifying the physicochemical characteristics of the subgroups, and then determining the polymer descriptor based on the subgroup descriptors for each polymer.

[0017] Preferably, the polymer descriptor refers to at least one of the following: compositional descriptor, counting descriptor, structural fragment list, fingerprint, graph invariant, 3D descriptor, and / or high-dimensional descriptor representing parameters quantifying physicochemical characteristics of a subgroup of polymers. In a preferred embodiment, the polymer descriptor refers to a 3D descriptor, specifically a quantum chemical descriptor. In particular, the quantum chemical descriptor is preferably determined for an embodiment utilizing polymerized monomers as subgroups. In general, the polymer descriptor is derived from the subgroup descriptor, and therefore the subgroup descriptor may also refer to the same descriptors as mentioned above. In the following, the expected descriptors are defined in more detail. Also, in this case, the defined descriptor may directly refer to the polymer descriptor or the subgroup descriptor.

[0018] The compositional descriptor may refer to any of the following: potential, average molecular weight, polydispersity, charge, spin, boiling point, melting point, enthalpy of fusion, dissociation constant, Hansen parameters, protic, polar and dispersive contributions, Abraham parameters, retention index, TPSA, receptor binding constant, Michaelis-Menten constant, inhibitor constant, mutagenicity, LD50, bioconcentration, toxicity, biodegradation profile, and viscosity.

[0019] The count descriptor refers to any of the following: atomic electronegativity, sum of atomic polarizabilities, amount of component, amount ratio of component, number of atoms and non-hydrogen atoms, number of H, B, C, N, O, P, S, Hal and heavy atoms, number of H donor atoms and H acceptor atoms, number of bonds, number of non-H bonds or multiple bonds, double bonds, triple bonds and aromatic bonds, number of functional groups, ratio of functional groups, number of chemical sites, sum of bond orders, ratio of aromatics, number of rings or cycles, number of unpaired electrons, number of rotatable bonds, ratio of rotatable bonds, number of conformers.

[0020] The polymer descriptor, which refers to a list of structural fragment descriptors, may refer to at least one of a list of molecular weights, a list of functional groups, a list of chemical sites, a list of bonds, and a list of atoms. The fingerprint descriptors preferably include at least one of the following: MACCS keys, preferably in bit format or total format, Morgan Fingerprints and other circular fingerprints, preferably in bit format or total format, topological twists, atom pairs, infrared spectra and related spectra, fingerprint numbers, PubChem fingerprints, substructure fingerprints, and Klekota-Roth fingerprints. The graph invariance / topology index descriptors preferably include at least one of the following: topostructural indexes and topochemical indexes.

[0021] In a preferred embodiment, the polymer descriptors are 3D descriptors including at least one or more of the following: total atomic volume, average volume per atom, total atomic area, average area per atom, area of ​​all atoms, average area per atom, solvent accessible surface, dispersion energy, dielectric energy, H donor, H acceptor, polar and non-polar surface areas, atomically resolved H donor, H acceptor, polar and non-polar surface areas, shape, sphericity, dipole and higher electric moments, polarizability, dielectric energy, proticity, polar and non-polar surface areas, orbital energies and orbital gaps, ionization energy, electron affinity, hardness, electronegativity, electrophilicity, excitation energy and intensity, infrared and ultraviolet absorption bands, reactivity measurements, redox potential, bond reference point, partial charges, charge surface area, atomic orbital contributions, bond order, atomic radius. In particular, the polymer descriptor preferably refers to a 3D descriptor comprising at least one of the following: total volume of all atoms, average volume per atom, total area of ​​all atoms, average area per atom, solvent accessible surface, dispersion energy, dielectric energy, H donor, H acceptor, polar and / or non-polar surface area, atomically resolved H donor, H acceptor, polar and / or non-polar surface area, shape, sphericity, cone angle, polarizability, dielectric energy, protonity, polar and / or non-polar surface area, excitation energy and intensity, infrared and / or ultraviolet absorption bands, reactivity measurements, particle charge, and / or charge surface area. Preferably, the high-dimensional descriptors utilized may include at least one or more of the following: configurational partition function, solubility, vapor pressure, activity coefficient, diffusion coefficient, partition coefficient, surface activity, rotational constant, moment of inertia, radius of gyration, compositional drift of the polymer, density, viscosity, conformer-weighted volume and area, conformer-weighted H donor, H acceptor, protonity, polar and / or non-polar surface area, charge distribution, configurational dipole moment, molecular refraction. Preferably, high-dimensional descriptors are utilized that include at least one of the following: solubility, vapor pressure and activity coefficient, surface activity, conformer-weighted H donor, H acceptor, protonity, polar and non-polar surface area, and charge distribution.

[0022] The method further comprises providing a predictive model adapted to predict the technical application properties of the polymer based on the digital representation. The predictive model is a data-driven model. In particular, the term "data-driven" is used herein to emphasize that the model is primarily based on the respective data inputs and not, for example, on intuition, personal experience or knowledge. Preferably, the predictive model refers to a machine learning-based model based on known machine learning algorithms such as neural networks, regression models, classification algorithms, etc. For most applications, it has been found that regression models based on random forest, LASSO, ridge regression and MARS algorithms are particularly suitable, while for classification models, random forest and SVM algorithms are particularly suitable. In a preferred embodiment, a neural network algorithm is used. In particular, the predictive model is parameterized such that the technical application properties associated with the polymer can be predicted based on the polymer descriptors represented by the digital representation. In general, the predictive model can be parameterized during a training process in which polymer descriptors derived from parameters quantifying the physicochemical characteristics of a subgroup of training polymers are utilized together with the corresponding technical application properties of the training polymers. Based on such a training data set, the respective parameters of the data-driven model can be determined such that the predictive model is capable of determining technical application properties of polymers that are not part of the training data set.

[0023] Furthermore, the method includes determining the technical application properties based on the provided digital representation of the polymer and the predictive model. In particular, if the digital representation of the polymer includes a polymer descriptor, the polymer descriptor is provided as an input to the predictive model, which then provides the predicted technical application properties as an output. If the digital representation does not directly include a polymer descriptor, determining the technical application properties may include determining the polymer descriptor, for example, as described above. The so determined polymer descriptor may then be provided as an input to the predictive model. The predicted technical application properties may then be provided, for example, to an output unit or to a computing unit for further processing. Preferably, providing the technical application properties leads to further processing with the technical application properties. In such a case, providing as a separate step may be omitted and replaced by processing the technical application properties.

[0024] Preferably, the processing of the technical application characteristic comprises determining a control signal for controlling the production process based on the determined estimated technical application characteristic. The production process may refer to the production process of the polymer itself or to the production process of the product in which the polymer is utilized. For example, if the determined technical application characteristic refers to the ignition temperature of the polymer to be utilized in the production process, the generation of the control signal may comprise generating a control signal that ensures that the production process is operated such that the polymer is always below the determined ignition temperature. In a preferred embodiment, the control signal represents a machine executable synthetic specification of the polymer, in particular if the result of the comparison indicates that the determined technical application characteristic of the polymer is within a predefined range around the provided target technical application characteristic. Preferably, in this embodiment, a further feedback loop is provided. In particular, the results of utilizing the control signal, for example the results of the controlled production process, can be monitored and the results of the monitoring can be compared with the expected results determined in the predictive model. Based on this comparison, the predictive model can be retrained, if necessary. For example, if the control signals represent a machine-executable synthesis specification, the polymers produced based on the control signals can be automatically put under a monitoring process that measures, for example, the respective technical application properties of the polymers and optionally further characteristics such as the polymer structure. These measurements can then be compared with the expected results predicted by the predictive model. If the comparison shows a difference above a predefined threshold, the measurements can be used directly to maintain the predictive model.

[0025] Moreover, the process of processing technical application properties may refer to a step of selecting one or more polymers based on the respective determined technical application properties. For example, if the respective technical application properties have been determined for a number of potential polymers, the selecting step may include comparing the technical application properties of the different polymers with predefined selection criteria and selecting those polymers whose predicted technical application properties meet these criteria. Specifically, in one embodiment, the method includes receiving target technical application properties for the polymers, comparing the received target technical application properties with the predicted technical application properties, and providing a control signal in response to the comparison. The control signal may refer to any signal that allows further control of the technical system. For example, the control signal may be adapted to control an interface to provide a comparison result on the interface. In a preferred embodiment, the comparison refers to a validation of the target technical application properties, the validation being positive if the predicted technical application properties fall within a predefined range around the target application properties. In this case, the control signal may be adapted to simply control a user interface to provide an indication of a positive or negative validation result. However, preferably, the control signal refers to a recipe, i.e. a synthesis specification, of one or more polymers that meet the particular target properties, i.e. the validation was positive. A recipe, i.e. a synthesis specification, is generally defined as an instruction on how a polymer can be synthesized. In particular, a recipe comprises starting materials and respective parameters for polymerization from the starting materials. Preferably, the control signal comprises the recipe in a form that directly enables the automatic control of the respective industrial system or work equipment for producing the polymer. In particular, it is preferred that the control signal represents a machine executable synthesis specification of the polymer if the result of the comparison indicates that the determined technical application property is within a predefined range around the target technical application property.

[0026] In a preferred embodiment, the method further comprises providing a synthesis specification as a digital representation of the polymer and determining polymer descriptors from the synthesis specification. In particular, the synthesis specification, i.e. recipe, comprises information about the polymer synthesis of the polymer, e.g., about the starting materials and about the process by which the respective starting materials are covalently linked to form the polymer chains or networks of the polymer. The method then comprises determining polymer descriptors from the synthesis specification. In particular, subgroups can be determined from the synthesis specification and then polymer descriptors can be determined based on the subgroup descriptors of the subgroups, e.g., from a database or by using known descriptor determination algorithms. In a preferred embodiment, further from the synthesis specification, the catalyst and / or non-reactive process components to be used are determined. In this case, this information is preferably also used together with the polymer descriptors by a predictive model for predicting the technical application properties. Preferably, a descriptor is also determined for the catalyst and / or non-reactive process components, and the respective descriptor is also used to determine the descriptors of the polymer. Preferably, the descriptors of the catalyst and / or non-reactive process components refer to the amounts of the respective components, e.g., molar mass, molar percentage, etc., and are considered to determine the polymer descriptor for the polymer.

[0027] In a preferred embodiment, the method comprises: i) determining polymerized monomers of the polymer from the digital representation, in particular from a synthetic specification of the polymer, ii) determining one or more subgroup descriptors for the determined polymerized monomers, and iii) determining a polymer descriptor based on the determined subgroup descriptors. Preferably, the determined subgroup descriptors and polymer descriptors refer to quantum chemical descriptors.

[0028] In a preferred embodiment, determining the polymer descriptor from the synthesis specification comprises, for example, identifying the type and amount of subgroups based on the synthesis specification as a descriptor of the subgroups, and determining the polymer descriptor based on the identified type and amount of subgroups. In general, the type of subgroup may refer to a predefined type or class associated with a specific physicochemical characteristic of the subgroups, i.e. the descriptor, and thus of the polymer comprising these subgroups. However, since the general physicochemical characteristic of the polymer, and therefore the polymer descriptor, may also depend on the amount of subgroups present in the polymer, the amount may also be taken into account. In a preferred embodiment, determining the type and amount of subgroups takes into account information provided by the synthesis specification representing the type of polymerization. Information on the type of polymerization available may, for example, refer to whether the polymerization refers to polycondensation, polyaddition, radical polymerization, cationic polymerization, anionic polymerization, or concerted chain polymerization. Preferably, for each type of polymerization, rules are predefined that can be applied to determine the subgroups of the polymer. For example, rules can be predefined that determine which functional groups and / or chemical moieties of the monomers in the synthesis specification react with which functional groups and / or chemical moieties of the synthesized polymer, and in what order of priority. The rules can be based, for example, on kinetic considerations. Based on the number and type of polymerization functional groups and / or chemical moieties, subgroups can be determined and the number and type of subgroups can be calculated.

[0029] In one embodiment, determining the amount of the subgroups includes determining the amount of at least one of amides, esters, thioesters, carbonates, ethers, amines, ureas, urethanes, thiourethanes, isocyanurates, biurets, allophanates, acetals, Michael adducts, radical polymerized double bonds, siloxanes, silanes, silazanes, phosphazene groups, and residual amines, aldehydes, ketones, epoxides, aziridines, isocyanates, alcohols, thiols, carboxylic acids, acyl halides, α,β unsaturated carbonyl groups, α,β unsaturated carboxyls, and double bond groups in the polymer based on synthesis specifications.

[0030] In a further aspect, a computer implemented method for predicting technical application properties for a polymer is presented, the method comprising the steps of: a) providing a synthesis specification of a polymer as a digital representation via a user interface; b) deriving polymer descriptors from the synthesis specification by i) identifying subgroups of polymers in the synthesis specification, ii) determining parameters quantifying physicochemical characteristics of the subgroups of polymers, and iii) determining polymer descriptors based on the subgroup parameters; c) determining and providing predicted technical application properties of the polymer based on the polymer descriptors as a digital representation using the computer implemented method described above; and d) providing the predicted technical application properties to a user via a user interface.

[0031] In particular, the determined parameters quantifying the physicochemical characteristics of the polymer subgroups can be considered as subgroup descriptors, i.e., subgroup descriptors. Respective examples for the identification of subgroups and the determination of subgroup descriptors have already been given above. For example, subgroups can be identified based on predefined rules from information provided by the synthesis specification. In addition, subgroup descriptors can be determined, for example, using known molecular simulation models or using respective databases in which subgroup descriptors are already stored for a number of expected subgroups.

[0032] In a further aspect, a computer-implemented training method for training a data-driven based predictive model to parameterize the predictive model is presented, the training method comprising the steps of: i) providing training data including: a) polymer descriptors for each of the training polymers, the polymer descriptors representing parameters quantifying physicochemical characteristics of a subgroup of the respective training polymers; and b) technical application properties associated with each training polymer; ii) providing a trainable data-driven based predictive model; iii) training the provided data-driven based predictive model based on the provided training data, such that the trained predictive model is adapted to predict technical application properties of the polymers based on the polymer descriptors; and iv) providing a trained predictive model.

[0033] In a further aspect, a computer-implemented optimization method for optimizing a synthesis specification for a polymer is provided, the method comprising: a) receiving, via an interface, i) a synthesis specification for a polymer to be optimized, ii) a target technical application property for which the synthesis specification is optimized, and iii) one or more optimization constraints, the constraints representing constraints on realization of the synthesis specification; and b) optimizing the synthesis specification for the polymer with respect to the target technical application property and the optimization constraints, the optimization comprising I) i) identifying subgroups of polymers in the synthesis specification, ii) determining parameters quantifying physicochemical characteristics of the subgroups of polymers, and iii) performing a polymorphism based on the parameters of the subgroups. II) utilizing the above-mentioned method by providing the polymer descriptor as a digital representation and determining predicted technical application properties of the polymer based on the digital representation; III) comparing the predicted technical application properties with target technical application properties and determining the synthetic specification as an optimal synthetic specification if the predicted technical application properties are within a predetermined range around the target application properties, and ii) iteratively optimizing to determine a modified synthetic specification taking into account the optimization constraints, and IV) generating a control signal based on the optimal synthetic specification.

[0034] In general, constraints may refer to all constraints that limit feasible solutions. For example, constraints may refer to constraints provided by the technical details of the industrial plant that will produce the optimized polymer, such as the expected temperature range, the flow rates of the semi-finished products, especially the reactants, the availability of certain materials such as catalysts required to produce a certain polymer, etc. For optimization, any known iterative optimization method may be used that allows an efficient search of the polymer space limited by the respective constraints for polymers that satisfy the technical application properties. To determine the modified synthesis specification, for example, Bayesian optimization methods may be used. In addition, rules may be used to add or remove monomers from the synthesis specification. In general, determining the modified synthesis specification may refer to a purely automated process. Or it may refer to an interactive process in which a user interacts with a computer system, for example by approving or marking the modified synthesis specification.

[0035] In a further aspect of the invention, a method for generating control signals, i.e. control data, suitable for manufacturing / producing a polymer is provided, the method comprising: a) receiving, via an interface: i) a potential target synthetic specification for a potential target polymer; ii) a target technological application property to be satisfied by the target synthetic specification; and iii) one or more constraints, the constraints representing constraints on realization of the target synthetic specification; and b) determining a target synthetic specification for the target polymer relative to the target technological application property and the constraints, the determining comprising: I) i) identifying a subgroup of potential target polymers in the potential target synthetic specification; ii) determining parameters quantifying physicochemical characteristics of the subgroup of potential target polymers; and iii) determining a target synthetic specification for the potential target polymer based on the parameters of the subgroup. II) utilizing the above-mentioned method by providing the polymer descriptor as a digital representation and determining and providing predicted technical application properties of the potential target polymer based on the digital representation; III) comparing the predicted technical application properties with the target technical application properties to: i) determine the potential target synthetic specification as the target synthetic specification if the predicted technical application property is within a predetermined range centered on the target application property, and ii) iterate this determination taking into account the constraints to determine a revised potential target synthetic specification; and IV) generating a control signal based on the target synthetic specification.

[0036] In a further aspect of the present invention, an apparatus for predicting technical application properties relating to a polymer based on a digital representation of the polymer is presented, the apparatus comprising: a) a digital representation providing unit for providing a digital representation of the polymer representative of polymer descriptors, the polymer descriptors representing parameters quantifying physicochemical characteristics of a subgroup of polymers; b) a predictive model providing unit for providing a predictive model adapted to predict the technical application properties of the polymer based on the digital representation, the predictive model being a data-driven model parameterized to predict the technical application properties associated with the polymer based on the polymer descriptors represented by the digital representation; c) a characterization unit for determining the technical application properties based on the provided digital representation of the polymer and the predictive model; and d) an output unit for providing the technical application properties.

[0037] In a further aspect of the present invention, an interface system for predicting technical application properties for polymers is presented, the system comprising: a) an interface adapted to receive a synthesis specification of a polymer; b) a derivation unit for deriving polymer descriptors for the polymer from the synthesis specification by i) identifying subgroups of polymers in the synthesis specification, ii) determining parameters quantifying physicochemical characteristics of the subgroups of polymers, and iii) determining a descriptor for the polymer based on the parameters of the subgroups; and c) a connection unit for providing the polymer descriptors as digital representations to the above mentioned device, for determining and providing predicted technical application properties for the polymer based on the digital representations, and for receiving the predicted technical application properties from the device in order to provide the technical application properties to a user via the interface.

[0038] In a further aspect of the present invention, a training apparatus for training a data-driven based predictive model to parameterize the predictive model is presented, the training apparatus comprising: i) a training data providing unit for providing training data including: a) digital representations of a plurality of training polymers, the digital representations comprising polymer descriptors for each of the training polymers, the polymer descriptors representing parameters quantifying physicochemical characteristics of a subgroup of the respective training polymers; and b) technical application properties associated with each training polymer; ii) a model providing unit for providing a trainable data-driven based predictive model; iii) a training unit for training the provided data-driven based predictive model based on the provided training data, such that the trained predictive model is adapted to predict technical application properties of the polymers based on the digital representations; and iv) a training model providing unit for providing a trained predictive model.

[0039] In a further aspect of the present invention, an optimization system for optimizing a synthesis specification of a polymer is presented, the system comprising: a) an interface adapted to receive i) a synthesis specification of a polymer to be optimized, ii) a target technological application property for which the synthesis specification is optimized, and iii) one or more optimization constraints, the constraints representing constraints on the realization of the synthesis specification; and b) an optimization unit for optimizing the synthesis specification of the polymer with respect to the target technological application property and the optimization constraints, the optimization comprising I) i) identifying subgroups of polymers in the synthesis specification, ii) determining parameters quantifying physicochemical characteristics of the subgroups of polymers, and iii) optimizing the subgroups based on the parameters of the subgroups. II) deriving a digital representation of the polymer from the synthesis specification by determining a descriptor of the polymer based on the synthesis specification; II) providing the digital representation to the above-mentioned device to determine and provide predicted technical application properties of the polymer based on the digital representation; III) comparing the predicted technical application properties with the target technical application properties and i) determining the synthesis specification as an optimal synthesis specification if the predicted technical application properties are within a predefined range around the target application properties, and ii) determining a modified synthesis specification by iterative optimization taking into account the optimization constraints if the predicted technical application properties are outside the predefined range around the target application properties; and IV) a control signal generation unit for generating a control signal based on the optimal synthesis specification.

[0040] In a further aspect of the present invention there is provided a computer program product for predicting technical application properties of a polymer, comprising program code means for causing an apparatus as above described to carry out a method as above described.

[0041] In a further aspect of the invention there is provided a computer program product for training a machine learning based predictive model, the computer program product comprising program code means for causing a training device as described above to perform the training method as described above.

[0042] In a further aspect of the present invention, a computer-implemented method for determining technical application properties for a polymer based on a digital representation of the polymer is presented, the method comprising: a) providing a digital representation of the polymer, the digital representation being representative of or associated with physicochemical characteristics of a subgroup of polymers; b) providing a predictive model adapted to determine the technical application properties of the polymer based on the digital representation, the predictive model being a data-driven model parameterized such that it is adapted to predict the technical application properties associated with the polymer based on the physicochemical characteristics exhibited by the digital representation; c) determining the technical application properties based on the provided digital representation of the polymer and the predictive model; and d) providing the technical application properties.

[0043] In a further aspect of the invention, a system is presented comprising: i) a control signal comprising a synthesis specification for a polymer, the synthesis specification indicating one or more ingredients for producing the polymer, the control signal being generated according to the method described above; and ii) one or more ingredients indicated by the synthesis specification in the control signal. Generally, the control signal may be embodied in or refer to control data and / or a control file, for example the control signal may be provided in a JSON format.

[0044] In a further aspect of the invention, the use of a control signal generated according to the method described above for controlling a production process, in particular a production process involving the production of polymers, is presented.

[0045] In a further aspect of the invention, a control signal is provided, the control signal being generated according to the method described above. Preferably, the control signal comprises a machine executable synthesis specification for producing a polymer.

[0046] It is to be understood that the above-mentioned method, the above-mentioned device and the above-mentioned computer program product have similar and / or identical preferred embodiments, in particular as defined in the dependent claims. In addition, furthermore, the above-mentioned training method, the above-mentioned training device and the above-mentioned training computer program product have similar and / or preferred embodiments, in particular as defined in the dependent claims.

[0047] It shall be understood that a preferred embodiment of the invention can also be any combination of the dependent claims or the above-mentioned embodiments with the respective independent claim.

[0048] These and other aspects of the invention will be apparent from and elucidated with reference to the embodiment(s) described hereinafter. [Brief description of the drawings]

[0049] [Figure 1] 1 shows a schematic and exemplary embodiment of a system comprising an apparatus for predicting technical application properties of a polymer. [Diagram 2] 1 shows, diagrammatically and exemplarily, a flow chart of a method for predicting technical application properties of a polymer. [Diagram 3] 1 shows, diagrammatically and exemplarily, a flow chart of a method for training a predictive model for predicting technical application properties of a polymer. [Figure 4] 1 shows, diagrammatically and exemplarily, a flow chart of a detailed embodiment of a method for predicting technical application properties of a polymer. [Diagram 5] 1 shows, generally and exemplarily, a flow chart of a detailed embodiment of a method for determining a target polymer. [Figure 6] 1 shows, generally and exemplarily, a flow chart of a detailed embodiment of a method for determining a target polymer. [Figure 7] 1 shows a schematic and exemplary block diagram of a system architecture of a system and apparatus for predicting technical application properties of polymers. [Figure 8]1 shows a schematic and exemplary block diagram of a system architecture of a system and apparatus for predicting technical application properties of polymers. [Figure 9] 1 shows a schematic and exemplary block diagram of a system architecture of a system and apparatus for predicting technical application properties of polymers. [Figure 10] 1 shows, in a schematic and exemplary manner, polymerized monomers derived from a number of monomers provided for polycondensation. [Figure 11] 1 shows, in a schematic and exemplary manner, polymerized monomers derived from a number of monomers provided for polyaddition. [Figure 12] Schematically and illustratively, polymerization monomers derived from the number of monomers provided for vinylic polymerization. [Figure 13] 1 shows a schematic and exemplary representation of polymerized monomers derived from the number of monomers provided for block-wise polyalkoxylation. [Figure 14] 1 shows a schematic and exemplary representation of polymerized monomers derived from a number of monomers provided for the PolyMichael addition. [Figure 15] 1 shows a schematic and exemplary polymerized monomer of polysiloxane. [Figure 16] A possible user interface is shown diagrammatically and exemplarily. [Figure 17] A possible user interface is shown diagrammatically and exemplarily. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0050] Detailed Description of the Preferred Embodiments 1 shows, in a schematic and exemplary manner, an embodiment of a system 100 comprising an apparatus 110 for predicting a technical application property of a polymer based on a digital representation of the polymer, further comprising a training apparatus 130 for training a predictive model used in the apparatus 110, a database 140 in which the results of the prediction of the technical application property of the polymer may be stored, and a production system 120 for producing a product that may be controlled using the predicted technical application property.

[0051] The apparatus 110 comprises a digital representation providing unit 111, a predictive model providing unit 112, a characterization unit 113 and an output unit 114. In addition, the apparatus comprises a control unit 115, which may optionally be adapted to provide control signals for controlling the production process of the production system 120. The digital representation providing unit 111 is adapted to provide a digital representation representing polymer descriptors of a polymer whose technical application properties have to be predicted. The digital representation providing unit 111 may for example refer to an input unit, through which a user can input the respective digital representation. In addition, the input unit may refer to or be part of a user interface allowing a user to interact with the apparatus 110 and / or the database 140. However, the digital representation providing unit 111 may also refer to or be communicatively coupled to a storage unit in which the digital representations of the polymers are already stored. In general, the digital representation may directly include polymer descriptors representing parameters quantifying physicochemical characteristics of the respective polymer subgroups. However, instead of directly providing the polymer descriptors, it is also possible to provide a synthesis specification of the polymers. In this case, it is preferred that the digital representation providing unit 111 is further adapted to determine the polymer descriptor from the composition specification. In particular, it is preferred that the digital representation providing unit 111 is adapted to identify the type and amount of polymer subgroups from the composition specification and to determine the polymer descriptor based on the identified type and amount of subgroups. In particular, the digital representation providing unit 111 may be adapted to determine a respective subgroup descriptor for each of the identified subgroups, for example by accessing a database in which respective descriptors are stored for a plurality of most relevant subgroups. The polymer descriptor may then be determined based on the subgroup descriptors of the subgroups, and preferably further based on the determined amount and type of subgroups, for example by taking a weighted average of the subgroup descriptors of the subgroups. More details regarding the determination of the polymer descriptors are provided in the detailed examples below.The digital representation providing unit 111 is then adapted to provide the digital representation comprising the polymer descriptors to, for example, a characterization unit 113 .

[0052] The predictive model providing unit 112 is adapted to provide a predictive model adapted to predict technical application properties of a polymer based on the digital representation. Here again, the predictive model providing unit 112 may comprise or refer to an input unit, via which the predictive model may be received, for example by a user inputting or indicating which predictive model should be used. In addition, the predictive model providing unit 112 may refer to or be communicatively coupled to a storage unit in which the predictive model is already stored. In a preferred embodiment, the predictive model providing unit 112 further refers to a selection unit, which is adapted, for example, to access a storage unit in which a plurality of predictive models are stored and to select a respective predictive model for providing for each application case. For example, different predictive models for different polymer types and / or for predicted technical application properties may be stored in such a storage unit, and the predictive model providing unit 112 acting as a selection unit may be adapted, for example, to select a corresponding predictive model based on a user input indicating the respective polymer and the respective predicted technical application properties. In addition, the selection unit may also be adapted to allow the user to select the respective predictive model or to assist the user in the selection of the respective predictive model suitable for the intended application. For such user interaction, for example, the user interface described above can be utilized.

[0053] The predictive model is a data-driven model that is parameterized so that the technical application properties associated with the polymer can be predicted based on the digital representation, in particular based on the polymer descriptors that represent the physicochemical characteristics of the subgroup. In a preferred embodiment, the data-driven model refers to a machine learning model, for example a regression model-based algorithm or a classifier model-based algorithm. The regression model-based algorithm can be based on any of the following algorithms: neural network algorithm, LASSO algorithm, ridge regression algorithm, MARS algorithm, and random forest algorithm. The classifier model-based algorithm can be based on any of the following algorithms: random forest algorithm and SVM algorithm. The inventors have found that for most applications, the random forest-based algorithm and the MARS-based algorithm are particularly suitable.

[0054] The predictive model can be trained, for example, using a training device 130. In particular, the training device 130 comprises a training data providing unit 131 for providing training data for training the data-driven based predictive model. The training data includes a) polymer descriptors for a plurality of training polymers, and b) one or more technical application properties associated with each training polymer. Preferably, in the training data, the technical application property provided for each training polymer refers to the same technical application property, depending on which technical application property the predictive model is to be trained for. For example, the technical application property may refer to a fire point, such that the technical application property provided for each training polymer includes a value of the fire point for each of the training polymers. If multiple technical application properties are to be predicted by the predictive model, two or more respective technical application properties are provided for the training polymers in the training data. In general, the training data may be designed to cover a given application space of the predictive model to be trained. For example, the training data may be designed to cover a given polymer type for a given application. Known methods of designing and optimizing training data may be utilized for a given application space, such that the application space is sufficiently covered by the training data and random outliers are avoided.

[0055] Further, the training device 130 comprises a model providing unit 132 adapted to provide a data-driven based trainable predictive model, e.g., a predictive model including parameters that can be set during a training process to train the predictive model. For example, the trainable predictive model can be already stored in a storage unit, and the model providing unit 132 can access the storage unit to provide the trainable predictive model. In addition, the training device 130 comprises a training unit 133 for training the provided data-driven based predictive model based on the provided training data. In particular, training may refer to varying parameters of the predictive model based on the respective training data until the predictive model is adapted to predict the technical application properties of the polymer based on the digital representation. In general, any known training algorithm for training a data-driven model, in particular a machine learning based model, can be utilized. Preferably, during the training of the predictive model, the polymer's descriptors that have the most significant impact on the respective properties are also determined, and then the model is trained based on these most influential descriptors. To determine these most influential descriptors, for example, a cluster analysis or PCA analysis tool can be utilized. Specifically, the descriptors can be used to determine an application space of the training data, where the application space is defined by the polymer's descriptors and property data. Then, the determination of the most influential descriptors can be performed as a dimensionality reduction of the application space. Then, an algorithm can be applied to optimize the training data in the application space, for example, to cover as much of the application space as possible with the field training data.

[0056] The training device 130 then comprises a trained model providing unit 134 adapted to provide the trained predictive model, for example to a storage unit, in which the trained predictive models for different technical application properties or different types of polymers, respectively, are stored. However, the trained model providing unit 134 can also be adapted to provide the trained predictive model directly, for example to the predictive model providing unit 112 of the device 110.

[0057] In all cases, the predictive model providing unit 112 is then adapted to provide a suitable trained predictive model to the characterization unit 113. The characterization unit 113 may then use the predictive model and the provided digital representation to determine technical application properties. In particular, the characterization unit 113 may be adapted to use the polymer descriptors represented by the digital representation as input to a predictive model, which has been trained as already described above to provide as output a prediction of the technical application properties for which the predictive model was trained. An output unit 114, for example a display, may then be adapted to output the predicted technical application properties. However, the output unit 114 may additionally or instead be adapted to provide the predicted technical application properties to a database 140 in order to store for future use the polymers associated with the predicted technical application properties. In particular, if technical application properties have already been determined for different polymers and stored, for example, in a storage unit, i.e. the database 140, the output unit 114 may be adapted to select the respective polymers based on predefined criteria for the technical application properties. Then, the output unit 114 may be adapted to provide and / or output the selected polymer and its technical application properties, which is particularly suitable when a user is searching for a polymer having specific characteristics regarding one or more technical application properties among a plurality of candidate polymers.

[0058] Optionally, the apparatus 110 may comprise a control unit 115 adapted to provide control signals for controlling the production process of the production system 120 based on the predicted technological application properties. In particular, the control unit 115 is preferably adapted to receive target technological application properties of the polymer, compare the received target technological application properties with the predicted technological application properties and provide control signals in response to the comparison, preferably providing control signals indicative of the use or production of the polymer whose technological application properties are predicted. In addition, if the result of the comparison indicates that the determined technological application properties are within a predefined range around the target technological application properties, the control signal may represent a machine executable synthesis specification of the polymer whose technological application properties are predicted. However, the control unit 115 may also be adapted to control the production process of another product based on the predicted technological application properties, for example to provide control signals representing a machine executable synthesis specification for another product that utilizes or includes the respective polymer. For example, if a polymer utilized during the synthesis of another product is predicted to include a fire point having a particular value, the controlling unit 115 may be adapted to output control signals that ensure that during the synthesis of the product, the production system 120 is always operated below the respective fire point.

[0059] Fig. 2 shows, in a schematic and exemplary manner, a flow chart of a method for predicting technical application properties of a polymer based on a digital representation of the polymer. The method 200 comprises, in a first step 210, providing a digital representation representing polymer descriptors of the polymer. In particular, providing the digital representation in this step may follow the principles described above with respect to the digital representation providing unit 111. Furthermore, in step 220, a predictive model is provided, adapted to predict technical application properties of the polymer based on the digital representation. As already described in more detail above, the predictive model is a data-driven model parameterized so that the technical application properties associated with the polymer can be predicted based on the polymer descriptors representing parameters quantifying the physicochemical characteristics of the subgroups. In general, steps 210 and 220 can be performed in any order or even simultaneously. In a next step 230, the technical application properties are determined based on the provided digital representation of the polymer and on the predictive model. The technical application properties can then be provided in step 240, for example for outputting the technical application properties on a display. Optionally, the method may further comprise, in step 250, generating a control signal making it possible to control the production process of a product, for example a polymer or a product comprising a polymer, as already described in more detail above.

[0060] FIG. 3 shows, in a schematic and exemplary manner, a flow chart of a method for training a data-driven based predictive model, for example as utilized in the method 200 discussed with respect to FIG. 2. In general, the method 300 may be implemented by, for example, a respective unit of the training device 130 as described with respect to FIG. 1. The method 300 comprises a step 310 of providing training data for training the data-driven based predictive model. The training data comprises a) polymer descriptors of a plurality of training polymers, and b) technical application properties associated with each training polymer. In particular, the training data may be provided according to the principles described above with respect to the training data providing unit 131 described with respect to FIG. 1. The method further comprises a step 320 of providing a data-driven based trainable predictive model, for example a machine learning based predictive model such as a neural network. In general, steps 310 and 320 may be performed in any order or even simultaneously. The method 300 then further comprises a step 330 of training the provided data-driven based predictive model based on the provided training data, for example by varying parameters in the data-driven based trainable predictive model such that the trained predictive model is adapted to predict technical application properties of the polymer based on the digital representation of the polymer. In step 340, the trained predictive model may then be provided, for example by storing the trained predictive model in a storage device or by directly providing the trained predictive model to the device 130, as described with respect to FIG.

[0061] In the following, more detailed preferred embodiments of the above-mentioned method and the corresponding device are described below. In particular, as described above, the method can be used to predict technical application properties of polymers belonging to at least the following groups: a) mechanical properties, such as adhesion, tensile strength, stiffness, hardness, shrinkage, elongation, tear, tear strength, elastic rebound, compressibility, abrasion, flow, morphology, tactile properties, stress at break, elongation at break, particle size distribution, and degree of packing; b) optical properties, such as color, turbidity, opacity, clarity, reflection, appearance, absorbance, scattering, color intensity; density, viscosity, K value, molar weight, dispersity / molar mass distribution, particle size distribution, solubility, partition coefficient, interfacial properties, surface tension, dispersibility, storage stability, odour, separation, coagulation, electrical conductivity, electrical capacity, surface area, flow time, vapour pressure, VOC, solids content, hygroscopicity, magnetism, miscibility, thixotropy, phase transition properties, glass transition temperature, corrosion inhibition, solvent separation, coagulation, self-heating, shock sensitivity, drying weight loss, reaction angle, electrostatic charge, minimum film formation temperature, charge density, dart drop, melt volume fraction, flowability, tear propagation resistance, seal strength, permeability, d) chemical properties such as functional group count, atom type count, functional group density, atom type count, chemical resistance, reaction timing, demoulding time, growth, hard / soft segment content, crystallinity, reaction temperature, reaction pressure, decomposition, pyrolysis, photolysis, acidity, pKa, pH, carbon footprint, production costs, waste formation, moisture / water content, flammability, burning rate, autoignition, flash point, generation of flammable gases, reaction to fire, deflagration rate, residual monomer count, generation of by-products, degree of polymerization, salt content, temperature resistance, oxidation properties, reduction properties, reactivity, ash content, non-volatile matter content, stability, chelating ability, calorific value, saponification value, e) biological properties such as biodegradability, biological resistance, toxicity, biotransformation, ecotoxicity, sensitization, bacterial count, enzyme activity, distribution in the environment, bioaccumulation.

[0062] Therefore, the method has applications in several technical fields, such as agricultural polymers, coatings, dispersions, structural polymers, e.g. polymer foams for thermal and acoustic insulation, shoes, automotive applications.

[0063] An exemplary embodiment of the method may consist of the steps described below. A schematic and exemplary flow chart of an exemplary embodiment of the method 400 is shown in FIG. 4. In a first step 410, a digital representation of a polymer is provided. The digital representation may directly include the polymer descriptor, in which case the steps up to step 450 shown in FIG. 4 may be omitted. However, in many cases the polymer descriptor must first be determined based on the provided digital representation, in such cases for example with reference to a recipe for the synthesis of the polymer or with reference to a chemical representation of the polymer showing the chemical components and bonds in the polymer. In this step 410, the digital representation may include any one or more of the following information: amounts of monomeric components; non-monomeric components such as initiators, fillers, additives; reaction conditions such as temperature, vessel, pressure, stirring speed; condition profiles, e.g. temperature profile, pH value, solvent; feed profile; type of polymerization, e.g. radical, cation, anion, polycondensation, polyaddition, polyether formation; post-treatment, such as amounts of components, conditions, and temperature and feed profile; type of post-treatment, e.g. radical, cation, anion, polycondensation, polyaddition, polyether formation; chemical information about compounds such as mixtures, connectivity of non-polymerizable pure compounds, composition of polymerizable pure compounds based on subgroups, connectivity of monomers associated with subgroups in polymerizable pure components; for block copolymers, information about the blocks in which each monomer and reactive prepolymer is incorporated; for structured / layered materials and compounds, information about the phase / layer in which each component is included. If such information is not directly provided by the digital representation, in optional step 420, reactive components and subgroups can also be derived from the digital representation, e.g. from a recipe.

[0064] If the information provided indicates the presence of a mixture, in a next step the mixture is decomposed into its pure components and each polymer component is treated as an input polymer. In addition, the polymer components may also be converted to mol % if necessary.

[0065] In the next step 430, the polymerizable components can be converted into subgroups, e.g. repeating units, and the subgroups are determined as different types. For example, the polymerizable subgroups can be determined based on the connectivity information of the non-polymerizable pure compounds, e.g. by using SMARTS, also via a KNIME workflow. Also, the connectivity information of all expected subgroups can be derived from the connectivity information of the non-polymerizable pure compounds, e.g. by using reaction SMARTS, also via a KNIME workflow.

[0066] After the subgroups and their types are determined, in step 440, the type of the descriptor to be used may be provided. However, the descriptor may also be determined without first selecting the type of subgroup. In order to reduce the computational resources for the method, in step 441, it is preferably determined whether the subgroup descriptor associated with each type of subgroup is already stored in the database, for example whether an entry for a subgroup with the same connectivity information is already stored in the database. In this case, the respective associated subgroup descriptor may be directly downloaded, for example in step 444. If the determined type of subgroup is not stored in the database, the subgroup descriptor associated with each type of subgroup may be determined, for example in step 442. For example, a 3D structure of the subgroups of each type may be derived based on the connectivity information, and the computation of the subgroup descriptor may be started automatically, for example using a computer cluster, or existing machine learning predictions may be used as the subgroup descriptor. In general, if computation for new subgroups is required, after the computation is completed, the results are preferably stored in the database in step 443. Optionally, further subgroup descriptors may be provided from subgroup topological analysis, quantum chemical calculations, molecular mechanics calculations, coarse-grained methods, finite element calculations, and dynamics simulations. In particular, polymer reaction engineering approaches may be used to derive subgroup descriptors that allow for taking into account the polymer microstructure.

[0067] In step 431, the amount of subgroups, i.e. the amount of each type of subgroup, is determined, for example based on the recipe information provided for the polymer. For example, the amount can be determined by counting the amount of polymerizable groups per polymerizable component, optionally including prepolymer. In this case, information on the polymerizable groups can be derived from the non-polymerizable components, and such determined amount is optionally added to the count of the number of unpolymerized polymerizable groups of the subgroups for the polymerizable component, based on the composition of the polymerizable component, to determine the resulting amount. Furthermore, it is preferred that the amount of polymerizable groups originating from the agent used for post-treatment after polymerization is removed from the resulting amount.

[0068] In the following, some preferred exemplary rules and schemes for determining subgroups and subgroup descriptors for different cases from a digital representation, for example from a recipe of a polymer, are described. In an embodiment where the provided digital representation, for example a recipe, indicates that the polymer is produced using polycondensation and polyaddition, the amount of reactive polymerizable groups is preferably determined based on kinetic considerations. Preferably, all functional groups of the monomers that can react in polycondensation or polyaddition are counted. These functional groups are referred to below as polymerizable groups. Preferably, the polymerizable groups are classified into two groups, namely nucleophilic groups and electrophilic groups. Then, all polymerizable amine groups, alcohol groups and thiol groups are determined to belong to the nucleophilic groups. All carboxylic acids, α,β unsaturated carboxylic acids, (methyl- and ethyl-) esters, anhydrides, carbonates and isocyanate groups are determined to belong to the electrophilic groups. Furthermore, it can be assumed that during polymerization, amine groups react more rapidly than alcohol groups, and both of these react more rapidly than thiol groups. Preferably, the reactive groups are all polymerizable groups that are polymerized in the final polymer. Preferably, the non-reactive groups are all polymerizable groups that are not polymerized in the final polymer (also called residual functional groups). Therefore, when the number of polymerizable amine groups is greater than the number of electrophilic polymerizable groups, it is preferred that the number of polymerizable amine groups is divided into non-reactive and reactive parts so that the number of reactive amine groups is equal to the number of electrophilic groups. In addition, the amount of non-reactive alcohol groups can then be determined to be equal to the number of polymerizable alcohol groups. The amount of non-reactive thiol groups can then be determined to be equal to the number of polymerizable thiol groups. When the number of polymerizable amine groups is less than or equal to the number of polymerizable electrophilic groups, it is preferred to classify all polymerizable amine groups as reactive amine groups. When the number of electrophilic groups is greater than the number of polymerizable amine groups, it is preferred to subtract the number of polymerizable amine groups from the number of electrophilic groups. This difference can then refer to the amount of remaining electrophilic groups in this case.

[0069] When the number of polymerizable alcohol groups is greater than the remaining electrophilic polymerizable groups, it is preferred that the number of polymerizable alcohol groups is divided into non-reactive and reactive portions so that the number of reactive alcohol groups is equal to the number of remaining electrophilic groups.The amount of non-reactive thiol groups is then equal to the number of polymerizable thiol groups.When the number of polymerizable alcohol groups is less than or equal to the electrophilic polymerizable groups, all of the polymerizable alcohol groups can be classified as reactive alcohol groups.

[0070] If the number of remaining electrophilic groups is greater than the number of polymerizable alcohol groups, it is preferred to classify all polymerizable alcohol groups as reactive alcohol groups. It is preferred to subtract the number of polymerizable alcohol groups from the number of remaining electrophilic groups. This difference then refers in this case to the new amount of remaining electrophilic groups. If the number of polymerizable thiol groups is greater than the new remaining polymerizable electrophilic groups, it is preferred to divide the number of polymerizable thiol groups into non-reactive and reactive parts so that the number of reactive thiol groups is equal to the new number of remaining electrophilic groups. If the number of polymerizable thiol groups is less than or equal to the electrophilic polymerizable groups, all polymerizable thiol groups can be classified as reactive alcohol groups.

[0071] If the number of remaining electrophilic groups is greater than the number of polymerizable thiol groups, then it is preferred to subtract the number of polymerizable thiol groups from the number of remaining electrophilic groups, in which case the difference refers to the amount of non-reactive electrophilic groups.

[0072] Based on the above, for example, the amount of amide groups, ester groups, thioester groups, urea groups, urethane groups, and thiourethane groups can be determined as described below. For example, the ratio N of reacted amine groups can be determined from the amount of all reactive nucleophilic groups. The ratio O of reacted alcohol groups can be determined from the amount of all reactive nucleophilic groups. The ratio S of reacted thiol groups can be derived from the amount of all reactive nucleophilic groups. The amount of reactive carboxyl groups can then be determined as the sum of reactive carboxylic acid groups, α,β unsaturated carboxylic acid groups, (methyl- and ethyl-) esters, anhydrides, and carbonate groups. The amount of reacted amide groups is then determined as a value equal to the amount of reactive carboxyl groups multiplied by the ratio N. In addition, the amount of reactive carboxyl groups is determined as a value equal to the amount of reactive ester groups multiplied by the ratio O. The amount of reacted thioester groups is determined as a value equal to the amount of reactive carboxyl groups multiplied by the ratio S. Additionally, the amount of reacted urea groups can be determined as equal to the amount of reactive isocyanate groups multiplied by the ratio N. The amount of reacted urethane groups can be determined as equal to the amount of reactive isocyanate groups multiplied by the ratio O. The amount of reacted thiourethane groups can be determined as equal to the amount of reactive isocyanate groups multiplied by the ratio S.

[0073] In general, if the recipe information provided indicates a prepolymer, then subgroups can also be determined for the prepolymer, e.g., as described above, and can be combined with the determined amounts of subgroups determined for the polymer. In addition, amounts of subgroups having the same connectivity and derived from the same monomer can be combined. If a post-treatment chemical is defined, polymerizable groups derived from the post-treatment chemical can be combined with the polymerizable group.

[0074] In step 432, such determined amount of subgroups may be provided and stored, for example, in a database. Before further processing the determined amount of subgroups, subgroups that are completely represented by other subgroups may be removed. In addition, subgroups with the same connectivity may be merged.

[0075] Optionally, subgroups of the derived quantities can be used for further interpretation of the polymer composition, for example, the total number of polymerizable functional groups, such as double bonds, amine groups, alcohol groups, thiol groups, carboxylic acid groups, isocyanate groups, epoxide groups, and formed functional groups, such as amide groups, ester groups, thioester groups, urea groups, urethane groups, thiourethane groups, ether groups, can be determined.Also, molar weighted total number of polymerizable functional groups, mass weighted total number of polymerizable functional groups, total number of remaining functional groups, e.g. double bonds, amine groups, alcohol groups, thiol groups, carboxylic acid groups, isocyanate groups, epoxide groups, molar weighted total number of remaining functional groups, mass weighted total number of remaining functional groups, sum of all remaining functional groups, ratio between functional groups after polymerization, number of crosslinks in the polymer, optionally mass weighted, mole fraction of crosslinks in the polymer, average number of atoms per subgroup, optionally per weight, average number of non-H atoms per subgroup, optionally per weight, average number of bonds per subgroup, optionally per weight, average number of bonds between non-H atoms per subgroup, optionally per weight, average number of rotors per subgroup, optionally per weight, average number of rotors between non-H atoms per subgroup, optionally per weight, average number of rings per subgroup, optionally per weight, average number of rings per subgroup, optionally per weight, The average polar surface area of ​​the polymer, the average refractive index per subgroup, optionally by weight, the total number of blocks, the molar size of the first block, the molar size of the last block, the HLB value of the polymer, optionally using the area-weighted HLB value, the HLB value of the block with the minimum HLB value, optionally using the area-weighted HLB value, the HLB value of the block with the maximum HLB value, optionally using the area-weighted HLB value, the HLB value of the first block, optionally using the area-weighted HLB value, the HLB value of the last block, optionally using the area-weighted HLB value, the mass of the first block, the mass of the last block, the area of ​​the block with the minimum HLB value, the area of ​​the block with the maximum HLB value, the difference in the HLB values ​​of the blocks, optionally using the area-weighted HLB value, the hydrophilic area of ​​the polymer, the lipophilic area of ​​the polymer, the number of arms in a ring-opening polymerization, or the arm length in a ring-opening polymerization may be determined.

[0076] Using the determined amount and type of subgroups and the associated subgroup descriptors, a polymeric descriptor may be calculated in step 450. For example, the polymeric descriptor may be determined by one or more of a molar weighted average, e.g., arithmetic mean, harmonic mean, logarithmic average, a mass weighted average, e.g., arithmetic mean, harmonic mean, logarithmic average, a volume weighted average, e.g., arithmetic mean, harmonic mean, logarithmic average, a surface area weighted average, e.g., arithmetic mean, harmonic mean, logarithmic average, of the associated descriptors of the subgroups. Additionally, the polymer descriptor may be determined by determining one or more of the molar weighted standard deviation, mass weighted standard deviation, volume weighted standard deviation, surface area weighted standard deviation, molar weighted maximum, mass weighted maximum, volume weighted maximum, surface area weighted maximum, molar weighted minimum, mass weighted minimum, volume weighted minimum, surface area weighted minimum, molar weighted sum, mass weighted sum, volume weighted sum, surface area weighted sum, and maximum difference from the associated subgroup descriptor.

[0077] In step 460, the derived or provided polymer descriptors can then be provided to a trained predictive model to predict application properties. In general, differently trained predictive models can also be utilized, as already mentioned above when predicting different polymer properties. However, the predictive model can also be adapted to predict multiple application properties. The predictive model can be trained based on an automatic statistical pre-processing of the training data, in particular of the training polymer descriptors, for example using feature engineering. For example, feature engineering can include first determining a number of different polymer descriptors for the polymer, for example based on subgroup descriptors of subgroups, and pre-selecting from this number of descriptors those that have a certain probability of being associated with the property. To identify groups of descriptors, a cluster analysis is preferably performed based on the relevant descriptors. Such groups make it possible to select only one of the members of the group, i.e. only one descriptor of the group, to represent the entire group of descriptors. Thus, based on the cluster analysis, the number of relevant descriptors can be further reduced. Based on the remaining descriptors, an application space is determined and optimized. The application of the trained predictive model is determined by the space spanned by the training data forming the application space. This space can be optimized, for example, by modifying the training data to cover the application space, removing strong outliers, adding training data to parts of the space that are not yet covered, etc. A predictive model is trained based on the optimized training data. A predictive model can generally refer to sparse, e.g., spline, LASSO regression, PLS, non-sparse, e.g., ridge regression, tree methods, kernel-based methods, statistical learning models for relating polymer descriptors to application properties of interest. In addition, the predictive model can further provide a reliability assessment of the prediction depending on the respective predictive model used. Then, in step 470, the predicted technical application properties can be provided to a user, for example, via a user interface.

[0078] The above-mentioned method can then be optionally deployed in the polymer database in the form of a data-driven model, which allows the composition of the desired polymer to be defined in a flexible and customer-oriented manner. For example, the database can be provided with a computer-implemented program that can provide as input information, i.e. as a digital representation, the assignment of components to specific polymer blocks, for example with respect to the amount of monomers and prepolymers, the type of polymerization, for example vinylic polymerization or polycondensation, and / or block copolymers, polymer composites, or polymers with composition shifts. Based on this input information, the method can be adapted to derive subgroups from user input, for example as described above. Furthermore, subgroup descriptors for the subgroups can be determined, in particular downloaded from the respective database. Polymer descriptors can then be derived based on the subgroup descriptors and the amount of the subgroups. In addition, additional polymer descriptors can also be calculated from the composition information. A prediction of one or more polymer properties can then be determined utilizing a machine learning-based predictive model pre-trained based on the polymer descriptors. The property predictions can then be provided to the front-end of the polymer database application.

[0079] In further applications, embodiments of the methods described above can be utilized for virtual screening using, for example, one of configuration optimization, Pareto optimization, low-dimensional visualization of screened recipes, feature selection for maximum applicability region of the model, and applicability region checker for new recipes at the descriptor level.

[0080] The advantages offered by utilizing the above-mentioned method are described below: Many polymer properties are based on the chemical nature of the polymer. It is therefore advantageous to use descriptors that reflect the chemical nature of the polymer. These are, for example, quantum chemical and topological descriptors derived from subgroups. Often, polymers are made from prepolymers, for which there are often many different grades, each of which is only included in very few polymer samples in the historical data set. This limits the performance of statistical models based only on recipe information. The subgroup-based method overcomes this limitation, since the prepolymers are also decomposed into subgroups. This is useful, since the prepolymers used often contain different amounts of similar subgroups, for example ethylene oxide and propylene oxide based subgroups for polyalkoxylates. As a result, the polymers in the historical data set contain similar subgroups, but the polymers contain different prepolymers. In addition, the expression of the polymer composition in terms of subgroups makes it possible to determine the chemical changes during polymerization. This information can be used as an additional source of descriptors. These additional descriptors can be advantageous for determining the applied properties using the above-mentioned method, depending on the application. Moreover, the above-described exemplary methods for deriving subgroups from other information, such as pure component connectivity information, are consistent for different monomers and different types of polymerization. This makes it possible to cover copolymers as well as mixtures of subgroups from different types of polymerization. This is advantageous, for example, when prepolymers such as polyalkoxylates are used in polyaddition or polycondensation.

[0081] Preferably, the polymer descriptors utilized in the above-mentioned methods are derived from quantum chemical calculations using solvation procedures. Quantum chemical calculations scale very poorly to system size, making calculations impractical for polymers or shorter monomer sequences. This obstacle is overcome by the above-mentioned methods, which involve cutting the polymer into subgroups, preferably at non-polarized bonds. The resulting subgroups have dimensions similar to the monomers, and the descriptors can be calculated using quantum chemical methods.

[0082] In particular, the above-mentioned method allows more polymer descriptors, and in particular more informative, i.e. more descriptive, polymer descriptors to be used as the basis of a predictive model for determining technical application properties. This in particular allows for a descriptor selection that allows for example to determine particularly suitable polymer descriptors for the prediction of technical application properties. This descriptor selection already provides a generalization capability during the training of the machine learning predictive model. This leads to a wider applicability domain and robustness of the predictive model. In addition, since the predictive model is based on polymer descriptors derived from subgroups, the predictive model has not learned how the monomers behave after polymerization. The predictive model is therefore more robust with respect to predicting polymers containing novel monomers that were not included in the training data. In addition, the prediction of polymer application properties using the above-mentioned method is computationally less expensive and therefore can be performed more quickly than generally known methods. This allows the optimization of polymer compositions, which requires providing predictions for hundreds or thousands of different polymers.

[0083] In the following, a more detailed preferred example of a preferred embodiment of the above-mentioned method and the corresponding device is described. A schematic and exemplary flow chart of an exemplary preferred embodiment of the method is provided by FIG. 5. In this exemplary embodiment, the method starts by requesting target values ​​for a target application, for example via a user interface. In addition, in a next step, the optimization is initialized by providing a potential target synthesis specification, i.e. a starting recipe. Optionally, in this process, constraints on the recipe, i.e. synthesis specification, can be taken into account, for example if the user provides such constraints. Constraints can refer, for example, to constraints on the production of the polymer, on the starting materials to be used to synthesize the polymer, etc. In addition, additional application conditions can be requested, which in particular represent further expected information for the target polymer to be realized. Based on the above steps, the optimization can be initialized to determine the target polymer, i.e. the target synthesis specification. In a first step of the optimization, for example, values ​​of the polymer descriptors can be derived from the provided starting recipe, i.e. from the provided potential target synthesis specification, as explained in detail with respect to FIG. 4. However, the derivation of the polymer descriptors can also refer to accessing a storage device in which the respective polymer descriptors of the respective potential target polymers are already stored. In addition, if the provided digital representation of the potential target composite specification already contains polymer descriptors, this step can also be omitted. Based on the required additional application conditions, a respective predictive model can be provided. Based on the provided predictive model and the digital representation of the potential target composite specification, a value of the target application of the potential target polymer can be provided. In the next step, it is determined whether the determined performance values, i.e. the determined technical application properties, meet the target values ​​within the given limits. If not, i.e. if this condition is not met, the formulation of the potential target composite specification is modified and a new target composite specification is determined, taking into account the previously provided constraints. Then, the iteration can start anew for the potential target composite specification.At some point, if the determined performance values ​​meet the target values ​​within their limits, i.e., if the respective conditions are realized, the potential target synthetic specifications are determined as target synthetic specifications and provided to a user or to a control unit, for example, to produce the respective determined target polymer.

[0084] FIG. 6 shows, in a schematic and exemplary manner, a further preferred embodiment of the above-mentioned method for determining a target synthesis specification with a given first target technical application property, in which in addition to the first target application property it is desirable for the target polymer to also realize a further second target value, i.e. the target technical application property. The additional target technical application property may refer to any technical application property. In general, specifically, the method follows the same principles as described above with reference to FIG. 5. However, due to the additional target value, additional conditions have to be met during the optimization. Therefore, in the following, only the main differences with respect to the above-mentioned method are pointed out. Specifically, in this preferred embodiment, the optimizer module not only performs the optimization over the first target value, i.e. over the first target application property, but also over a second target value. Preferably, also for this second target value, a predictive model is utilized, adapted to determine the value of the technical application property based on physicochemical parameters of the polymer. Thus, in addition to the method for a second target application as described above, a second predictive model is provided, which makes it possible to determine the application property value based on physicochemical parameters of the polymer for the second target application. The second predictive model can be based on, for example, the same algorithm as the first predictive model, only trained with a different data set to determine another property of the polymer. Then, the comparison refers not only to determining whether the determined first application property meets the target application property within limits, but also to determining whether the determined second application property meets the second target application property within limits. A predefined rule can be used to determine when to continue the iteration. That is, a new formula is provided as a new potential target composite specification, and a potential target composite specification is determined for that condition as the target composite specification. For example, the user can predetermine weighting values ​​for weighting which conditions must be met to what degree. For example, it may be more important for the user that the first technical application property is met, while other target application properties are less important.In this case, the constraints on which the second target application property may be satisfied may be set broader, or the weighting of satisfying this condition may be reduced. Then, at some point in the iteration, if the condition is determined to be met and satisfies the predetermined rule, each potential target synthesis specification may be determined as a target synthesis specification and may be provided as an output to a user or may be utilized to generate a control file for producing each target polymer.

[0085] FIG. 7 shows a block diagram of an exemplary system architecture of an automated laboratory system 1000 for synthesizing polymers, which includes a laboratory equipment control device 1102, a network 1150, and a synthesis specification module, i.e., recipe, module 1100 / 1110, and a client device 1108. The automated laboratory system includes a laboratory equipment control device layer 1152 as part of the laboratory equipment control device 1102, and a synthesis specification module layer 1154 associated with the synthesis specification module, and a remote control or client layer 1156 associated with the client device 1108. The laboratory equipment control device layer can be divided into several hierarchical layers, namely, a hardware layer, a middleware layer, and an interface layer. The hardware layer specifically relates to hardware resources, e.g., sensors and actuators, for controlling the synthesis of polymers. The middleware relates to any of the known middleware for laboratory or plant synthesis operations. One example is LABS / QM, which provides various abstractions over hardware, networks, and operating systems, such as low-level device control and message passing. The communication layer relates to communication protocols, one of which is REST, which can be implemented over different transport protocols (i.e. UDP, TCP, Telemetry) and allows the exchange of messages between the laboratory equipment control device and the laboratory equipment device. Such a software architecture allows to control and monitor the laboratory equipment without interacting with the hardware.

[0086] The synthesis specification module layer 1154 may include a mass storage layer, a computing layer, and an interface layer. The mass storage layer is configured to provide mass storage for a data-driven predictive model for providing a recipe, i.e., a synthesis specification, for a polymer based on technical application properties, as described in detail above. In particular, the functions performed by the device as described above may be provided as program code means stored in the mass storage device. Furthermore, the synthesis specifications for multiple polymers may be stored in the mass storage device. Such data may be stored in a structured database, such as a SQL database, or a distributed file system, such as HDFS, or a NoSQL database, such as HBase, MongoDB, etc. The computing layer may include an application layer that allows customizing the functionality provided by standard cloud services to perform computing processes based on target properties. Such functions may include determining a digital representation of a target polymer based on target technical application properties and a predictive model, generating a synthesis specification from the digital representation of the target polymer, and providing the synthesis specification as control data, i.e., a control signal, to a laboratory equipment control device.

[0087] The interface layer can implement web services, network interfaces as UDP or TCP, or web socket interfaces. For communication with laboratory equipment control devices, a REST API is implemented.

[0088] The client layer 1156 provides an interface to an end user. For the end user, the client layer 1156 can execute a client-side web application that provides an interface to the composite specification module layer 1154 or the laboratory equipment control device layer 1152. A UI may be provided to the user to select a target technical application characteristic and a target value for this characteristic, which may also include a range of technical application characteristics. In other embodiments, a UI may be provided to the user to select multiple technical application characteristics and respective values. The application may be configured for the user to remotely monitor and control the laboratory equipment control device and its operation. In other embodiments, the client device layer and the composite specification module layer may be integrated into one device. The alternatives described herein are merely for illustrative purposes and should not be considered limiting.

[0089] 8 illustrates a block diagram of an example system architecture of a system and apparatus for generating a predictive model for predicting technical application characteristics, a network 2150, a model generation module 2100 / 2110 that may be considered or comprises a training model apparatus, a synthesis specification module 1100 / 1110, and a client device 2108. The system for generating a predictive model includes a model generation module layer 2154 as part of the model generation module and a client layer 2156 associated with the client device 2108.

[0090] The model generation module layer 2154 may include a mass storage layer, a computing layer, and an interface layer. The storage layer is configured to provide mass storage for the data-driven predictive model as described above. Furthermore, the mass storage is configured to store the polymer synthesis specifications and the measured technical application properties. Such data may be stored in a structured database, such as a SQL database, or a distributed file system, such as HDFS, or a NoSQL database, such as HBase, MongoDB, etc. The computing layer may include an application layer that allows customizing the functionality provided by standard cloud services to perform computing processes for generating predictive models for predicting the properties of the polymer. Such functionality may include: receiving, for at least two previously measured polymers, their respective digital representations associated with synthetic specification measurement data of at least one technical application property for each of the at least two previously measured polymers; receiving a digital representation of at least one unmeasured polymer in the model generation module; training a model according to the training principles described above based on a similarity measure between the digital representations of the at least two previously measured polymers, the measurement data of at least one technical application property for each of the at least two previously measured polymers, and preferably the digital representation associated with the synthetic specification for each of the at least two previously measured polymers and the digital representation associated with the synthetic specification of the at least one unmeasured polymer; and providing a predictive model for the technical application property via an output interface. The model generation module layer may be configured to deploy the generated model and the synthetic specification database to the synthetic specification module layer. This may include storing the generated model and the synthetic specification database in a mass storage device associated with the synthetic specification module.

[0091] The model generation module layer may further be configured to determine from the synthesis specification a digital representation of the polymer associated with the synthesis specification. The digital representation may include a set of polymer descriptors and polymer descriptor values ​​associated with the synthesis specification for each measured polymer. One way to derive these polymer descriptors may be to apply the SMILES algorithm or any other principles already mentioned above. If a model is generated based on a digital representation derived from a recipe, the relationship between the synthesis specification and the descriptors may be stored in a mass storage device associated with the model generation module. In such a case, developing the model includes providing the relationship.

[0092] The interface layer may implement a web service, a network interface as UDP or TCP, or a web socket interface. In this example, a REST API is implemented for communication with the client device. The client layer 2156 provides access to a mass storage device, which includes a composition specification of polymers and at least one technical application property for at least two polymers. The client layer further provides an interface to an end user. For the end user, the client layer 2156 may execute a client-side web application that provides an interface to the model generation module layer 2154 or to a mass storage device associated with the client layer. The user may be provided with a UI for selecting the technical application property. The user may further be provided with a UI for selecting the composition specification data and the technical application property data associated with the composition specification data. The user interface may also provide an option to upload the selected data to the model generation module layer and, optionally, to start the model generation.

[0093] FIG. 9 shows an exemplary system 700 for producing a chemical product based on a synthesis specification generated according to the present invention. In this example, the system comprises a user interface 710 and a processor 720 associated with a control unit 740. The user interface 710 and the processor 720 can be associated with or implemented according to the above-mentioned principles, in particular adapted to execute a computer-implemented method for determining a target polymer and / or synthesis specification based on predicted technical application properties, as described above. The control unit 740 is configured, for example, to receive control data, i.e. control signals, generated according to the present invention as described above, in particular to receive control data generated based on a synthesis specification of a polymer including the target technical application properties. In this embodiment, the control data is provided from a database 730, but in other embodiments, the control data can also be provided from a server or any other computing unit for distributing data. The vessels 750, 752 each contain a component of a chemical product, e.g., a prepolymer, a catalyst, etc. Typically, there are three or more vessels, but in this embodiment only two are shown for illustrative purposes. Valves 760, 762 are associated with the vessels 750, 752. Valves 750 and 752 may be controlled to dose the appropriate amount of each component into reaction vessel 770 according to the synthesis specification. A motor 800 for agitator 780 may also be controlled by the control unit according to the synthesis specification. An optional heater 790 may also be controlled according to the synthesis specification. Finally, an outlet valve 810 in fluid communication with the reactor may be controlled by the control unit to provide chemical products to a vessel or test system 820.

[0094] Below, some examples of subgroups determined for a particular polymer are described in more detail with respect to Figs. 10-15. In particular, these below provided examples of derivation of subgroups are merely illustrative and utilize models and assumptions that provide suitably accurate predicted results. However, other models and assumptions can be used to derive the subgroups. In particular, utilizing a kinetic model to derive the subgroups allows for further increase in accuracy when determining the subgroups and can also improve the accuracy of the predicted results. Fig. 10 shows, for example, polymerized monomers derived from a number of monomers provided for polycondensation. In this example, 10 moles of adipic acid are polymerized with 5.5 moles of butane-1,4-diol and 5.5 moles of ethylene glycol. 10 moles of adipic acid monomer are polymerized to form 10 moles of adipic acid dimethyl ester, i.e., polymerized monomers of adipic acid are formed in the resulting polymer. The additional carbon atom of the polymerized monomer is taken from a monomer containing an alcohol group. Thus, the polymerized monomer of the monomer butane-1,4-diol in the polymer chain is ethane. In the case of ethylene glycol, the polymerized monomers in the polymer chain are completely represented by the polymerized adipic acid monomer (= adipic acid dimethyl ester), which is coded by a cross in the scheme. This polymerized monomer of ethylene glycol can be neglected for the calculation of the descriptors. According to the number of monomers provided, there is an excess of 2 moles of alcohol groups compared to the acid groups. Assuming a complete polymerization of adipic acid and equal reactivity of the two diols, i.e. butane-1,4-diol and ethylene glycol, 0.5 moles of butanediol and 0.5 moles of ethylene glycol remain unreacted. These unreacted monomers resemble the polymerized monomers at the end of the polymer chain. Besides this assumption, it is also possible to obtain a more realistic distribution of the polymerized monomers, for example using kinetic models.

[0095] FIG. 11 shows, for example, the polymerized monomers derived from the number of monomers provided for polyaddition. In this example, 10 moles of hexamethylene diisocyanate are polymerized with 6 moles of cyclohexane-1,4-diol, 3 moles of glycerol, and 2 moles of butane 1,4-diamine. In this example, it is assumed that amines react more favorably with isocyanates than with alcohols. In a first step, 2 moles of hexamethylene diisocyanate are polymerized to a polymerized monomer containing urea groups, in which the urea groups formed are N-substituted by methyl groups taken from the amine-containing compound. As a result, the monomer butane 1,4-diamine is polymerized to ethane as a polymerized monomer, since the original amine groups and two of the four carbon atoms of the monomer 1,4-butane diamine already originate from the polymerized monomer of hexamethylene diisocyanate. In a second step, the alcohol groups are polymerized with the remaining 8 moles of hexamethylene diisocyanate. There is an excess of alcohol groups compared to the number of isocyanate groups. Thus, 8 moles of hexamethylene diisocyanate are polymerized into 8 moles of urethane groups with polymerized monomers O-substituted by methyl groups. Assuming equal reactivity of cyclohexane-1,4-diol and glycerol, 4.57 moles of polymerized cyclohexane-1,4-diol are formed, which is represented by cyclohexane. In this special case, no carbon atoms are removed from the monomer, because such removal would change the ring size of cyclohexane-1,4-diol. We assume that the reactivity of all three alcohol groups of glycerol is similar, so that all three alcohol groups react virtually with isocyanate. For each alcohol group reacted, a methoxy group is removed from glycerol. As a result, 2.29 moles of polymerized glycerol can be completely represented by urethane with polymerized monomers of hexamethylene diisocyanate and can be ignored for the calculation of the descriptors. Due to the excess of alcohol groups, 1.43 moles of cyclohexane-1,4-diol and 0.71 moles of glycerol remain unreacted.These unreacted monomers resemble polymerized monomers at the ends of the polymer chain. In addition to these assumptions, a kinetic model can be used to obtain a more realistic distribution of polymerized monomers. In this example, 2.29 moles of polymerized glycerol are formed, as described above. This polymerized monomer has three reactive functional groups and acts as a crosslink in the final polymer. This information on the crosslinks can be used as a descriptor to distinguish between linear and crosslinked polymers.

[0096] FIG. 12 shows the polymerized monomers derived from the number of monomers provided for vinylic polymerization. In this example, 10.5 moles of methyl acrylate are polymerized with 3.7 moles of styrene. In this example, a complete conversion of the monomers during the polymerization is assumed. Thus, 10.5 of polymerized methyl acrylate and 3.7 of polymerized styrene are formed. The polymerized monomers can be defined in various ways. On the left side, the polymerized monomers are represented by molecular structures in which the reactive double bonds of the corresponding monomers are saturated by a hypothetical hydrogenation treatment (addition of two hydrogen atoms). For example, the monomer styrene can be represented by ethylbenzene as the polymerized monomer. On the right side, additional groups, e.g. methyl groups, are added, which resemble the electronic effect of the polymer chain on the polymerized monomer. However, these additional groups are preferably ignored by the descriptor calculation, e.g. by ignoring their contribution to the molecular surface area. Besides this assumption, it is also possible to obtain a more realistic distribution of the polymerized monomers using kinetic models.

[0097] Figure 13 shows the polymerized monomers derived from the number of monomers provided for the polyalkoxylation per block. In the first step, water is used as a model initiator to represent the hydroxy salt of the polyalkoxylation in which 4 moles of ethylene oxide are polymerized. The polymerized monomer of water is dimethyl ether. After the polymerization of 4 moles of ethylene oxide, two of them are located in the polymer chain and two are at the chain end. The polymerized monomer of ethylene oxide in the chain is likewise dimethyl ether. The polymerized monomer of ethylene oxide at the chain end is methanol. All the polymerized monomers contribute to the inner block of the final block copolymer. In the second step, 6 moles of propylene oxide react with the polymer from step 1. At this time, the polymerized monomers at the chain end from step 1 react with propylene oxide. As a result, these polymerized monomers at the chain end of step 1 are now located in the polymer chain, and the resulting polymerized monomer is again dimethyl ether. For 6 moles of propylene oxide, 4 of them form the polymerized monomer (methoxyethane) in the polymer chain and 2 moles form the polymerized monomer (ethanol) at the chain end. All the polymerized monomers derived from the monomer propylene oxide originate from the outer blocks of the block copolymer obtained after step 2. In the third step, the chain end modification of the block copolymer formed in the first two steps is carried out. This chain end modification is carried out via partial esterification with 0.6 moles of butyric acid, i.e., a condensation reaction with the loss of one water molecule per newly formed ester bond. The polymerized monomer of butyric acid after esterification is methyl butyrate. The additional oxygen and carbon atoms of this polymerized monomer are taken from the polymerized monomer containing an alcohol group, which is the polymerized monomer of propylene oxide at the chain end of the polymer obtained after step 2, i.e., ethanol. Correspondingly, those of the polymerized monomer (ethanol) that formed the ester by partial esterification are now converted to methane. Besides this assumption, it is possible to obtain a more realistic distribution of the polymerized monomers, also using a kinetic model.

[0098] Figure 14 shows the polymerized monomers derived from the number of monomers provided for the polyMichael addition. In this example, 5 moles of ethylene glycol diacrylate are polymerized with 4 moles of butane-1,4-dithiol. The Michael acceptor groups (acrylate groups) are in excess over the Michael donor groups (thiol groups). In this example, it is assumed that the 4 moles of butane-1,4-dithiol are fully converted during the polymerization. The resulting polymerized monomer of the monomer butane-1,4-dithiol is 1,4-bis(methylsulfanyl)butane. The additional carbon atom of this polymerized monomer is taken from the monomer containing the Michael acceptor group. As a result, the 4 moles of monomer ethylene glycol diacrylate that reacted form 4 moles of ethylene glycol diacetate as the polymerized monomer (loss of a carbon atom). The remaining excess 1 mole of ethylene glycol diacrylate monomer does not react. These unreacted monomers resemble polymerized monomers at the polymer chain end. Besides this assumption, it is also possible to use kinetic models to obtain a more realistic distribution of the polymerized monomers.

[0099] FIG. 15 shows the polymerized monomers of polysiloxane. In this example, 20 moles of dichlorodimethylsilane are polymerized with 2 moles of chlorotrimethylsilane and 21 moles of water (hydrolysis and subsequent polyaddition). In this example, a complete conversion of the monomers during polymerization is assumed. Thus, in this example, 20 polymerized monomers of dichlorodimethylsilane are formed in the polymer chain, as well as 2 moles of polymerized monomers of chlorotrimethylsilane (represented by hydroxytrimethylsilane) at the chain ends. Besides this assumption, it is possible to obtain a more realistic distribution of the polymerized monomers using kinetic models. In this example, it is not possible to define the polymerized monomers in the polymer chain such that the polymer is broken in a non-polarized homogeneous chemical bond. Therefore, additional groups (e.g., methyl and methoxy groups) are added that resemble the electronic effect of the polymer chain on the polymerized monomers. However, these additional groups must be ignored by the descriptor calculation (e.g., by ignoring their contribution to the molecular surface area). By using the methyl and methoxy groups as models of the polymer chain, the polymerized monomer of dichlorodimethylsilane is dimethoxydimethylsilane. Alternatively, small oligomers of dichlorodimethylsilane and chlorotrimethylsilane can be used as polymerized monomers.

[0100] 16 and 17 exemplarily and diagrammatically show possible user interfaces interfacing with a processor implementing one of the above-mentioned methods for determining application properties or target polymers, i.e. optimized polymers. FIG. 16 exemplarily shows a user interface for determining application properties of a polymer. In this example, an input screen is shown on the left. The input screen allows the definition of the object for which the technical application properties are to be determined and, optionally, the definition according to which the application method is to be measured. In this example, hardness and gloss must be determined. Furthermore, the input screen may allow providing a digital representation of the polymer for which the application properties are to be determined. In this case, the digital representation is defined by the polymer class being vinylic and further details regarding the type of monomers, as well as the associated blocks and amounts of the polymer. Then, based on this input, the target properties are predicted according to one of the above-mentioned methods. An exemplary output screen is shown on the right side of FIG. 16. In this example, the output screen shows the results of the prediction of two application properties.

[0101] FIG. 17 shows an exemplary user interface for determining optimized polymers, i.e. target polymers, with respective target technical application properties. In this example, an input screen is shown on the left. The input screen allows the definition of one or more target application properties. In this example, respective value ranges are provided for the hardness and gloss of the polymer. Furthermore, the input screen allows providing respective constraints for the target polymer. In this example, the target polymer must be constrained to the vinylic polymer class, and further constraints for the monomers are defined in the form of minimum and maximum values. Based on these input parameters, a respective target polymer that satisfies these parameters is determined according to one of the methods described above. An exemplary output screen is shown on the right of FIG. 17. In this example, the output screen provides details of the determined target polymer together with the respective determined values ​​for the application properties. Other variations to the disclosed embodiments can be understood and implemented by those skilled in the art upon studying the drawings, the disclosure, and the appended claims when implementing the claimed invention.

[0102] With respect to the processes and methods disclosed herein, the operations performed in the processes and methods may be performed in different orders. Furthermore, the operations outlined are provided only as examples, and some of the operations are optional and may be combined into fewer steps and operations, supplemented with further operations, or expanded with additional operations, without detracting from the essence of the embodiments of the present disclosure.

[0103] In the claims, the word "comprising" does not exclude other elements or steps, and the indefinite article "a" or "an" does not exclude a plurality.

[0104] A single unit or device may fulfill the functions of several items recited in the claims. The mere fact that certain measures are recited in mutually different dependent claims does not indicate that a combination of these measures cannot be used to advantage.

[0105] The procedures performed by one or more units or devices, such as providing polymer descriptors and predictive models, determining technical application properties, providing technical application properties, etc., may be performed by any other number of units or devices. These procedures may be implemented as program code means of a computer program and / or as dedicated hardware.

[0106] The computer program product may be stored on / distributed on a suitable medium, such as an optical storage medium or a solid-state medium, supplied together with or as part of other hardware, but may also be distributed in other forms, for example via the Internet or other wired or wireless telecommunications systems.

[0107] Any unit described herein may be a processing unit that is part of a classical computing system. The processing unit may include a general-purpose processor, or may include a field programmable gate array (FPGA), an application specific integrated circuit (ASIC), or any other dedicated circuit. Any memory may be physical system memory, and may be volatile, non-volatile, or a combination of both. The term "memory" may include computer readable storage media, such as non-volatile mass storage devices. If the computing system is distributed, the processing and / or storage capabilities may also be distributed. The computing system may include multiple structures as "executable components." The term "executable components" is a structure well understood in the computing arts, as structures that may be software, hardware, or a combination of both. For example, if implemented in software, those skilled in the art will understand that the structures of executable components may include software objects, routines, methods, etc. that may be executed on the computing system. This may include both executable components in the heap of the computing system, or executable components on computer readable storage media. The structure of the executable components may reside on a computer-readable medium such that when interpreted by one or more processors of a computing system, e.g., by processor threads, it causes the computing system to perform a function. Such structure may be directly computer readable by a processor, e.g., where the executable components are binary, or may be structured to be interpretable and / or compiled to generate such binary directly interpretable by a processor, e.g., whether in a single stage or multiple stages. In other examples, the structure may be hard-coded or hard-wired logic gates implemented exclusively or nearly exclusively in hardware, e.g., in a field programmable gate array (FPGA), application specific integrated circuit (ASIC), or other dedicated circuitry.Thus, the term "executable component" refers to structures well understood by those skilled in the computing arts, whether implemented in software, hardware, or a combination thereof. Any of the embodiments herein are described with reference to operations performed by one or more processing units of a computing system. When such operations are implemented in software, one or more processors direct the operation of the computing system in response to execution of the computer-executable instructions that make up the executable component. A computing system may also include communication channels that enable the computing system to communicate with other computing systems, for example, via a network. A "network" is defined as one or more data links that enable the transmission of electronic data between computing systems and / or modules and / or other electronic devices. When information is transferred or provided to a computing system via a network or another communications connection, for example, either hardwired, wireless, or a combination of hardwired and wireless, the computing system properly considers the connection to be a carrier medium. A carrier medium may include a network and / or data link that can be used to carry desired program code means in the form of computer-executable instructions or data structures and that can be accessed by a general-purpose computing system or a special-purpose computing system or a combination thereof. Although not all computing systems require a user interface, in some embodiments a computing system includes a user interface system for use in interfacing with a user. The user interface serves as an input or output mechanism to a user, for example via a display.

[0108] Those skilled in the art will appreciate that at least a portion of the present invention may be implemented in a networked computing environment having many types of computing system configurations, including personal computers, desktop computers, laptop computers, message processors, handheld devices, multiprocessor systems, microprocessor-based or programmable consumer electronics, network PCs, minicomputers, mainframe computers, cell phones, PDAs, pagers, routers, switches, data centers, wearable devices such as glasses, etc. The present invention may also be implemented in a distributed system environment in which tasks are performed together by local and remote computing systems that are linked, for example, via a network, either by hardwired data links, wireless data links, or a combination of hardwired and wireless data links. In a distributed system environment, program modules may be located in both local and remote memory storage devices.

[0109] Those skilled in the art will also appreciate that at least a portion of the present invention may be implemented in a cloud computing environment. A cloud computing environment may be distributed, but this is not required. When distributed, a cloud computing environment may be distributed internationally within an organization and / or may have components held across multiple organizations. In this specification and in the claims that follow, "cloud computing" is defined as a model that allows on-demand network access to a shared pool of configurable computing resources, e.g., networks, servers, storage, applications and services. The definition of "cloud computing" is not limited to any of the many other advantages that may be obtained when such a model is deployed. The computing system of the figures includes various components or functional blocks that may implement various embodiments disclosed herein, as described. The various components or functional blocks may be implemented on a local computing system or may be implemented on a distributed computing system that includes elements that reside in the cloud or implement aspects of cloud computing. The various components or functional blocks may be implemented as software, hardware, or a combination of software and hardware. The computing system shown in the figures may include more or fewer components than those shown in the figures, and some of the components may be combined as circumstances permit.

[0110] Any reference signs in the claims should not be construed as limiting the scope.

[0111] The present invention refers to an apparatus for predicting technical application properties for a polymer based on a digital representation of the polymer. A digital representation providing unit provides a digital representation of the polymer representing polymer descriptors. The polymer descriptors represent parameters quantifying physicochemical characteristics of a subgroup of the polymer. A predictive model providing unit provides a predictive model adapted to predict the technical application properties of the polymer based on the digital representation, the predictive model being a data-driven model parameterized to predict the technical application properties associated with the polymer based on the polymer descriptors represented by the digital representation. A characterization unit determines the technical application properties based on the provided digital representation of the polymer and the predictive model. An output unit provides the technical application properties.

Claims

1. 1. A computer-implemented method for predicting technical application properties for a polymer based on a digital representation of said polymer, said method (200) comprising: providing (210) a digital representation of the polymer representing polymer descriptors, the polymer descriptors indicating parameters quantifying physicochemical characteristics of a subgroup of the polymer; providing (220) a predictive model adapted to predict technical application properties of the polymer based on the digital representation, the predictive model being a data-driven model parameterized to be adapted to predict the technical application properties associated with the polymer based on the polymer descriptors indicated by the digital representation; determining (230) the technical application properties based on the provided digital representation of the polymer and the predictive model; and providing (240) said technical application characteristics.

2. The method of claim 1 , further comprising providing a synthetic specification of the polymer as a digital representation of the polymer, and determining the polymer descriptors from the synthetic specification.

3. 3. The method of claim 2, wherein the step of determining the polymer descriptor from the synthesis specification comprises the steps of identifying types and amounts of subgroups based on the synthesis specification, and determining the polymer descriptor based on the identified types and amounts of the subgroups.

4. The method of claim 3 , wherein the step of determining the type and amount of the subgroups takes into account information provided by the synthesis specification that describes the type of polymerization.

5. 10. The method of claim 1, further comprising receiving a target technical application property for a polymer, comparing the received target technical application property to a predicted technical application property, and providing a control signal in response to the comparison (250).

6. 6. The method of claim 5, wherein if the result of the comparison indicates that the determined technical application property is within a predetermined range centered on the target technical application property, the control signal represents a machine-executable synthesis specification for the polymer.

7. 1. A computer-implemented method for predicting technical application properties for a polymer, said method comprising: providing, via a user interface, a polymer synthesis specification as a digital representation; deriving the polymer descriptors from the synthesis specification by: i) identifying subgroups of the polymers in the synthesis specification; ii) determining parameters that quantify physicochemical characteristics of the subgroups of the polymers; and iii) determining descriptors for the polymers based on the parameters for the subgroups; Utilizing the computer-implemented method of claim 1 to determine and provide predicted technical application properties of the polymer based on the polymer descriptors as a digital representation; and providing the predicted technical application characteristics to a user via the user interface.

8. 1. A computer-implemented training method for training a data-driven based predictive model and parameterizing said predictive model, said training method (300) comprising: providing training data (310) including: a) polymer descriptors for each of the training polymers, the polymer descriptors representing parameters quantifying physicochemical characteristics of a subgroup of the respective training polymers; and b) technical application properties associated with each training polymer; Providing a trainable data-driven predictive model (320); training (330) the provided data-driven based predictive model based on the provided training data such that the trained predictive model is adapted to predict technical application properties of polymers based on polymer descriptors; and providing (340) the trained predictive model.

9. 1. A computer-implemented optimization method for optimizing a polymer synthesis specification, said method comprising: receiving, via an interface: i) a synthesis specification for a polymer to be optimized; ii) target technical application properties for which said synthesis specification is optimized; and iii) one or more optimization constraints, said optimization constraints representing constraints on the realization of said synthesis specification; optimizing the synthesis specifications of the polymer against the target technical application properties and the optimization constraints, the optimization comprising: deriving polymer descriptors for the polymers from the synthesis specification by: i) identifying subgroups of the polymers in the synthesis specification; ii) determining parameters that quantify physicochemical characteristics of the subgroups of the polymers; and iii) determining descriptors for the polymers based on the parameters of the subgroups; Utilizing the method of claim 1 by providing said polymer descriptors as digital representations and determining predicted technical application properties of said polymer based on said digital representations; an optimizing step, comprising: comparing the predicted technical application characteristics with the target technical application characteristics; i) determining the composite specification as the optimal composite specification if the predicted technical application characteristics are within a predetermined range centered on the target technical application characteristics; and ii) iterating the optimization to determine a modified composite specification taking into account the optimization constraints if the predicted technical application characteristics are outside the predetermined range centered on the target technical application characteristics; generating a control signal based on the optimal synthesis specification.

10. 1. An apparatus for predicting technical application properties for a polymer based on a digital representation of said polymer, said apparatus (110) comprising: a digital representation providing unit (111) for providing a digital representation of said polymer representing polymer descriptors, said polymer descriptors representing parameters quantifying physicochemical characteristics of a subgroup of said polymer; a predictive model providing unit (112) for providing a predictive model adapted to predict technical application properties of the polymer based on the digital representation, the predictive model being a data-driven model parameterized to predict the technical application properties associated with the polymer based on the polymer descriptors indicated by the digital representation; a characterization unit (113) for determining said technical application properties based on said digital representation of said polymer provided and said predictive model; and an output unit (114) for providing said technical application characteristics.

11. 1. An interface system for predicting technical application properties for polymers, said system comprising: an interface adapted to receive a polymer synthesis specification; a derivation unit for deriving polymer descriptors for said polymers from said synthesis specification by: i) identifying subgroups of said polymers in said synthesis specification; ii) determining parameters quantifying physicochemical characteristics of said subgroups of said polymers; and iii) determining descriptors for said polymers based on said parameters of said subgroups; a connection unit for providing the polymer descriptor as a digital representation to the device (110) of claim 10, for determining and providing predicted technical application properties of the polymer based on the digital representation, and for receiving the predicted technical application properties from the device (110) for providing the technical application properties to a user via the interface.

12. A training device for training a data-driven based predictive model and parameterizing said predictive model, said training device (130) comprising: a training data providing unit (131) for providing training data comprising: a) digital representations of a plurality of training polymers, said digital representations comprising polymer descriptors for each of said training polymers, said polymer descriptors representing parameters quantifying physicochemical characteristics of a subgroup of respective training polymers; and b) technical application properties associated with each training polymer; a model providing unit (132) for providing a trainable data-driven predictive model; a training unit (133) for training the provided data-driven based predictive model based on the provided training data, such that the trained predictive model is adapted to predict technical application properties of a polymer based on a digital representation; a training model providing unit (134) for providing the trained predictive model.

13. 1. An optimization system for optimizing a synthesis specification for a polymer, the system comprising: an interface adapted to receive: i) a synthetic specification of a polymer to be optimized; ii) a target technical application property for which said synthetic specification is optimized; and iii) one or more optimization constraints, said optimization constraints representing constraints on the realization of said synthetic specification; an optimization unit for optimizing the synthesis specifications of the polymer against the target technical application properties and the optimization constraints, the optimization comprising: deriving a digital representation of the polymer from the synthesis specification by: i) identifying subgroups of the polymer in the synthesis specification; ii) determining parameters that quantify physicochemical characteristics of the subgroups of the polymer; and iii) determining a descriptor of the polymer based on the parameters of the subgroups; providing said digital representation to an apparatus (110) according to claim 10, and determining and providing predicted technical application properties of said polymer based on said digital representation; an optimization unit including: comparing the predicted technical application characteristics with the target technical application characteristics; i) determining the composite specification as an optimal composite specification if the predicted technical application characteristics are within a predetermined range centered on the target technical application characteristics; and ii) iterating the optimization taking into account the optimization constraints to determine a modified composite specification if the predicted technical application characteristics are outside the predetermined range centered on the target technical application characteristics; a control signal generation unit for generating a control signal based on the optimal synthesis specification.

14. A computer program for predicting technical application properties for polymers, comprising program code means for causing an apparatus (110) according to claim 10 to perform a method (200) according to any one of claims 1 to 6.

15. 13. A computer program for training a machine learning based predictive model, the computer program comprising program code means for causing an apparatus (130) according to claim 12 to perform the method (300) according to claim 9.

16. 14. A system comprising: i) a control signal including a synthesis specification for a polymer, the synthesis specification representing one or more ingredients for producing the polymer, the control signal being generated according to any one of claims 5, 9 or 13; and ii) the one or more ingredients indicated by the synthesis specification in the control signal.

17. 14. Use of a control signal generated according to any one of claims 5, 9 or 13 for controlling a production process, in particular a production process involving the production of polymers.

18. A control signal generated in accordance with any one of claims 5, 9 or 13.

Images